Place floats after the first paragraph that mentions them

This means that most figures and tables should appear after their
first mention in the text. They might still float up because they're
floats, though, so further tweaking might be required.

thesis.lyx

@@ -41,6 +41,9 @@
 % (to avoid landscape figures breaking up text)
 \usepackage{afterpage}
 
+% Consider: force floats after placement in text
+% https://tex.stackexchange.com/questions/15706/force-floats-to-be-typeset-after-their-occurrence-in-the-source-text
+
 % This one breaks subfigs so it's disabled
 % https://tex.stackexchange.com/questions/65680/automatically-bold-first-sentence-of-a-floats-caption
 
@@ -3101,6 +3104,52 @@ literal "false"
  batch effect in the data.
 \end_layout
 
+\begin_layout Standard
+Due to an error in sample preparation, the RNA from the samples for days
+ 0 and 5 were sequenced using a different kit than those for days 1 and
+ 14.
+ This induced a substantial batch effect in the data due to differences
+ in sequencing biases between the two kits, and this batch effect is unfortunate
+ly confounded with the time point variable (Figure 
+\begin_inset CommandInset ref
+LatexCommand ref
+reference "fig:RNA-PCA-no-batchsub"
+plural "false"
+caps "false"
+noprefix "false"
+
+\end_inset
+
+).
+ To do the best possible analysis with this data, this batch effect was
+ subtracted out from the data using ComBat 
+\begin_inset CommandInset citation
+LatexCommand cite
+key "Johnson2007"
+literal "false"
+
+\end_inset
+
+, ignoring the time point variable due to the confounding with the batch
+ variable.
+ The result is a marked improvement, but the unavoidable confounding with
+ time point means that certain real patterns of gene expression will be
+ indistinguishable from the batch effect and subtracted out as a result.
+ Specifically, any 
+\begin_inset Quotes eld
+\end_inset
+
+zig-zag
+\begin_inset Quotes erd
+\end_inset
+
+ pattern, such as a gene whose expression goes up on day 1, down on day
+ 5, and back up again on day 14, will be attenuated or eliminated entirely.
+ In the context of a T-cell activation time course, it is unlikely that
+ many genes of interest will follow such an expression pattern, so this
+ loss was deemed an acceptable cost for correcting the batch effect.
+\end_layout
+
 \begin_layout Standard
 \begin_inset Float figure
 wide false
@@ -3233,49 +3282,43 @@ PCoA plots of RNA-seq data showing effect of batch correction.
 \end_layout
 
 \begin_layout Standard
-Due to an error in sample preparation, the RNA from the samples for days
- 0 and 5 were sequenced using a different kit than those for days 1 and
- 14.
- This induced a substantial batch effect in the data due to differences
- in sequencing biases between the two kits, and this batch effect is unfortunate
-ly confounded with the time point variable (Figure 
+However, removing the systematic component of the batch effect still leaves
+ the noise component.
+ The gene quantifications from the first batch are substantially noisier
+ than those in the second batch.
+ This analysis corrected for this by using 
+\begin_inset Flex Code
+status open
+
+\begin_layout Plain Layout
+limma
+\end_layout
+
+\end_inset
+
+'s sample weighting method to assign lower weights to the noisy samples
+ of batch 1 (Figure 
 \begin_inset CommandInset ref
 LatexCommand ref
-reference "fig:RNA-PCA-no-batchsub"
+reference "fig:RNA-seq-weights-vs-covars"
 plural "false"
 caps "false"
 noprefix "false"
 
 \end_inset
 
-).
- To do the best possible analysis with this data, this batch effect was
- subtracted out from the data using ComBat 
+) 
 \begin_inset CommandInset citation
 LatexCommand cite
-key "Johnson2007"
+key "Ritchie2006,Liu2015"
 literal "false"
 
 \end_inset
 
-, ignoring the time point variable due to the confounding with the batch
- variable.
- The result is a marked improvement, but the unavoidable confounding with
- time point means that certain real patterns of gene expression will be
- indistinguishable from the batch effect and subtracted out as a result.
- Specifically, any 
-\begin_inset Quotes eld
-\end_inset
-
-zig-zag
-\begin_inset Quotes erd
-\end_inset
-
- pattern, such as a gene whose expression goes up on day 1, down on day
- 5, and back up again on day 14, will be attenuated or eliminated entirely.
- In the context of a T-cell activation time course, it is unlikely that
- many genes of interest will follow such an expression pattern, so this
- loss was deemed an acceptable cost for correcting the batch effect.
+.
+ The resulting analysis gives an accurate assessment of statistical significance
+ for all comparisons, which unfortunately means a loss of statistical power
+ for comparisons involving samples in batch 1.
 \end_layout
 
 \begin_layout Standard
@@ -3332,36 +3375,6 @@ RNA-seq sample weights, grouped by experimental and technical covariates.
 
 \end_layout
 
-\begin_layout Standard
-However, removing the systematic component of the batch effect still leaves
- the noise component.
- The gene quantifications from the first batch are substantially noisier
- than those in the second batch.
- This analysis corrected for this by using 
-\begin_inset Flex Code
-status open
-
-\begin_layout Plain Layout
-limma
-\end_layout
-
-\end_inset
-
-'s sample weighting method to assign lower weights to the noisy samples
- of batch 1 
-\begin_inset CommandInset citation
-LatexCommand cite
-key "Ritchie2006,Liu2015"
-literal "false"
-
-\end_inset
-
-.
- The resulting analysis gives an accurate assessment of statistical significance
- for all comparisons, which unfortunately means a loss of statistical power
- for comparisons involving samples in batch 1.
-\end_layout
-
 \begin_layout Standard
 In any case, the 
 \begin_inset Flex Glossary Term
@@ -3490,62 +3503,28 @@ ChIP-seq differential modification analysis
 \end_layout
 
 \begin_layout Standard
-\begin_inset Float figure
-wide false
-sideways false
-status collapsed
-
-\begin_layout Plain Layout
-\align center
-\begin_inset Float figure
-wide false
-sideways false
+\begin_inset Flex TODO Note (inline)
 status open
 
 \begin_layout Plain Layout
-\align center
-\begin_inset Graphics
-	filename graphics/CD4-csaw/csaw/CCF-plots-noBL-PAGE2-CROP.pdf
-	lyxscale 50
-	height 40theight%
-	groupId ccf-subfig
-
-\end_inset
-
-
-\end_layout
-
-\begin_layout Plain Layout
-\begin_inset Caption Standard
-
-\begin_layout Plain Layout
-
-\series bold
-\begin_inset CommandInset label
-LatexCommand label
-name "fig:CCF-without-blacklist"
-
+Be consistent about use of 
+\begin_inset Quotes eld
 \end_inset
 
-Cross-correlation plots without removing blacklisted reads.
- 
-\series default
-Without blacklisting, many artifactual peaks are visible in the cross-correlatio
-ns of the ChIP-seq samples, and the peak at the true fragment size (147
-\begin_inset space ~
+differential binding
+\begin_inset Quotes erd
 \end_inset
 
-bp) is frequently overshadowed by the artifactual peak at the read length
- (100
-\begin_inset space ~
+ vs 
+\begin_inset Quotes eld
 \end_inset
 
-bp).
-\end_layout
-
+differential modification
+\begin_inset Quotes erd
 \end_inset
 
-
+ throughout this chapter.
+ The latter is usually preferred.
 \end_layout
 
 \end_inset
@@ -3553,249 +3532,73 @@ bp).
 
 \end_layout
 
-\begin_layout Plain Layout
-\align center
-\begin_inset Float figure
-wide false
-sideways false
+\begin_layout Standard
+Sequence reads were retrieved from 
+\begin_inset Flex Glossary Term
 status open
 
 \begin_layout Plain Layout
-\align center
-\begin_inset Graphics
-	filename graphics/CD4-csaw/csaw/CCF-plots-PAGE2-CROP.pdf
-	lyxscale 50
-	height 40theight%
-	groupId ccf-subfig
+SRA
+\end_layout
 
 \end_inset
 
+ 
+\begin_inset CommandInset citation
+LatexCommand cite
+key "Leinonen2011"
+literal "false"
 
-\end_layout
+\end_inset
 
-\begin_layout Plain Layout
-\begin_inset Caption Standard
+.
+ 
+\begin_inset Flex Glossary Term (Capital)
+status open
 
 \begin_layout Plain Layout
-
-\series bold
-\begin_inset CommandInset label
-LatexCommand label
-name "fig:CCF-with-blacklist"
+ChIP-seq
+\end_layout
 
 \end_inset
 
-Cross-correlation plots with blacklisted reads removed.
+ (and input) reads were aligned to GRCh38 genome assembly using Bowtie 2
+ 
+\begin_inset CommandInset citation
+LatexCommand cite
+key "Langmead2012,Schneider2017,gh-hg38-ref"
+literal "false"
 
-\series default
- After blacklisting, most ChIP-seq samples have clean-looking periodic cross-cor
-relation plots, with the largest peak around 147
-\begin_inset space ~
 \end_inset
 
-bp, the expected size for a fragment of DNA from a single nucleosome, and
- little to no peak at the read length, 100
-\begin_inset space ~
-\end_inset
+.
+ Artifact regions were annotated using a custom implementation of the 
+\begin_inset Flex Code
+status open
 
-bp.
+\begin_layout Plain Layout
+GreyListChIP
 \end_layout
 
 \end_inset
 
+ algorithm, and these 
+\begin_inset Quotes eld
+\end_inset
 
-\end_layout
-
+greylists
+\begin_inset Quotes erd
 \end_inset
 
+ were merged with the published 
+\begin_inset Flex Glossary Term
+status open
 
+\begin_layout Plain Layout
+ENCODE
 \end_layout
 
-\begin_layout Plain Layout
-\begin_inset Caption Standard
-
-\begin_layout Plain Layout
-\begin_inset Argument 1
-status collapsed
-
-\begin_layout Plain Layout
-Strand cross-correlation plots for ChIP-seq data, before and after blacklisting.
-\end_layout
-
-\end_inset
-
-
-\begin_inset CommandInset label
-LatexCommand label
-name "fig:CCF-master"
-
-\end_inset
-
-
-\series bold
-Strand cross-correlation plots for ChIP-seq data, before and after blacklisting.
-\end_layout
-
-\end_inset
-
-
-\end_layout
-
-\end_inset
-
-
-\end_layout
-
-\begin_layout Standard
-\begin_inset Note Note
-status open
-
-\begin_layout Plain Layout
-\begin_inset Float figure
-wide false
-sideways false
-status collapsed
-
-\begin_layout Plain Layout
-\align center
-\begin_inset Graphics
-	filename graphics/CD4-csaw/ChIP-seq/H3K4me2-sample-MAplot-bins-CROP.png
-	lyxscale 25
-	width 100col%
-	groupId colwidth-raster
-
-\end_inset
-
-
-\end_layout
-
-\begin_layout Plain Layout
-\begin_inset Caption Standard
-
-\begin_layout Plain Layout
-
-\series bold
-\begin_inset CommandInset label
-LatexCommand label
-name "fig:MA-plot-bigbins"
-
-\end_inset
-
-MA plot of H3K4me2 read counts in 10kb bins for two arbitrary samples.
-\end_layout
-
-\end_inset
-
-
-\end_layout
-
-\end_inset
-
-
-\end_layout
-
-\end_inset
-
-
-\end_layout
-
-\begin_layout Standard
-\begin_inset Flex TODO Note (inline)
-status open
-
-\begin_layout Plain Layout
-Be consistent about use of 
-\begin_inset Quotes eld
-\end_inset
-
-differential binding
-\begin_inset Quotes erd
-\end_inset
-
- vs 
-\begin_inset Quotes eld
-\end_inset
-
-differential modification
-\begin_inset Quotes erd
-\end_inset
-
- throughout this chapter.
- The latter is usually preferred.
-\end_layout
-
-\end_inset
-
-
-\end_layout
-
-\begin_layout Standard
-Sequence reads were retrieved from 
-\begin_inset Flex Glossary Term
-status open
-
-\begin_layout Plain Layout
-SRA
-\end_layout
-
-\end_inset
-
- 
-\begin_inset CommandInset citation
-LatexCommand cite
-key "Leinonen2011"
-literal "false"
-
-\end_inset
-
-.
- 
-\begin_inset Flex Glossary Term (Capital)
-status open
-
-\begin_layout Plain Layout
-ChIP-seq
-\end_layout
-
-\end_inset
-
- (and input) reads were aligned to GRCh38 genome assembly using Bowtie 2
- 
-\begin_inset CommandInset citation
-LatexCommand cite
-key "Langmead2012,Schneider2017,gh-hg38-ref"
-literal "false"
-
-\end_inset
-
-.
- Artifact regions were annotated using a custom implementation of the 
-\begin_inset Flex Code
-status open
-
-\begin_layout Plain Layout
-GreyListChIP
-\end_layout
-
-\end_inset
-
- algorithm, and these 
-\begin_inset Quotes eld
-\end_inset
-
-greylists
-\begin_inset Quotes erd
-\end_inset
-
- were merged with the published 
-\begin_inset Flex Glossary Term
-status open
-
-\begin_layout Plain Layout
-ENCODE
-\end_layout
-
-\end_inset
+\end_inset
 
  blacklists 
 \begin_inset CommandInset citation
@@ -3911,113 +3714,83 @@ literal "false"
 \end_layout
 
 \begin_layout Standard
-Promoters were defined by computing the distance from each annotated 
-\begin_inset Flex Glossary Term
+\begin_inset Float figure
+wide false
+sideways false
+status collapsed
+
+\begin_layout Plain Layout
+\align center
+\begin_inset Float figure
+wide false
+sideways false
 status open
 
 \begin_layout Plain Layout
-TSS
-\end_layout
+\align center
+\begin_inset Graphics
+	filename graphics/CD4-csaw/csaw/CCF-plots-noBL-PAGE2-CROP.pdf
+	lyxscale 50
+	height 40theight%
+	groupId ccf-subfig
 
 \end_inset
 
- to the nearest called peak and examining the distribution of distances,
- observing that peaks for each histone mark were enriched within a certain
- distance of the 
-\begin_inset Flex Glossary Term
-status open
 
-\begin_layout Plain Layout
-TSS
 \end_layout
 
+\begin_layout Plain Layout
+\begin_inset Caption Standard
+
+\begin_layout Plain Layout
+
+\series bold
+\begin_inset CommandInset label
+LatexCommand label
+name "fig:CCF-without-blacklist"
+
 \end_inset
 
-.
- For H3K4me2 and H3K4me3, this distance was about 1
+Cross-correlation plots without removing blacklisted reads.
+ 
+\series default
+Without blacklisting, many artifactual peaks are visible in the cross-correlatio
+ns of the ChIP-seq samples, and the peak at the true fragment size (147
 \begin_inset space ~
 \end_inset
 
-kb, while for H3K27me3 it was 2.5
+bp) is frequently overshadowed by the artifactual peak at the read length
+ (100
 \begin_inset space ~
 \end_inset
 
-kb.
- These distances were used as an 
-\begin_inset Quotes eld
-\end_inset
+bp).
+\end_layout
 
-effective promoter radius
-\begin_inset Quotes erd
 \end_inset
 
- for each mark.
- The promoter region for each gene was defined as the region of the genome
- within this distance upstream or downstream of the gene's annotated 
-\begin_inset Flex Glossary Term
-status open
 
-\begin_layout Plain Layout
-TSS
 \end_layout
 
 \end_inset
 
-.
- For genes with multiple annotated 
-\begin_inset Flex Glossary Term (pl)
-status open
 
-\begin_layout Plain Layout
-TSS
 \end_layout
 
-\end_inset
-
-, a promoter region was defined for each 
-\begin_inset Flex Glossary Term
-status open
-
-\begin_layout Plain Layout
-TSS
-\end_layout
-
-\end_inset
-
- individually, and any promoters that overlapped (due to multiple 
-\begin_inset Flex Glossary Term (pl)
-status open
-
-\begin_layout Plain Layout
-TSS
-\end_layout
-
-\end_inset
-
- being closer than 2 times the radius) were merged into one large promoter.
- Thus, some genes had multiple promoters defined, which were each analyzed
- separately for differential modification.
-\end_layout
-
-\begin_layout Standard
-\begin_inset Float figure
-wide false
-sideways false
-status collapsed
-
 \begin_layout Plain Layout
+\align center
 \begin_inset Float figure
 wide false
 sideways false
-status collapsed
+status open
 
 \begin_layout Plain Layout
 \align center
 \begin_inset Graphics
-	filename graphics/CD4-csaw/ChIP-seq/H3K4me2-PCA-raw-CROP.png
-	lyxscale 25
-	width 45col%
-	groupId pcoa-subfig
+	filename graphics/CD4-csaw/csaw/CCF-plots-PAGE2-CROP.pdf
+	lyxscale 50
+	height 40theight%
+	groupId ccf-subfig
 
 \end_inset
 
@@ -4032,37 +3805,30 @@ status collapsed
 \series bold
 \begin_inset CommandInset label
 LatexCommand label
-name "fig:PCoA-H3K4me2-bad"
+name "fig:CCF-with-blacklist"
 
 \end_inset
 
-H3K4me2, no correction
-\end_layout
+Cross-correlation plots with blacklisted reads removed.
 
+\series default
+ After blacklisting, most ChIP-seq samples have clean-looking periodic cross-cor
+relation plots, with the largest peak around 147
+\begin_inset space ~
 \end_inset
 
-
-\end_layout
-
+bp, the expected size for a fragment of DNA from a single nucleosome, and
+ little to no peak at the read length, 100
+\begin_inset space ~
 \end_inset
 
+bp.
+\end_layout
 
-\begin_inset space \hfill{}
 \end_inset
 
 
-\begin_inset Float figure
-wide false
-sideways false
-status collapsed
-
-\begin_layout Plain Layout
-\align center
-\begin_inset Graphics
-	filename graphics/CD4-csaw/ChIP-seq/H3K4me2-PCA-SVsub-CROP.png
-	lyxscale 25
-	width 45col%
-	groupId pcoa-subfig
+\end_layout
 
 \end_inset
 
@@ -4073,15 +3839,25 @@ status collapsed
 \begin_inset Caption Standard
 
 \begin_layout Plain Layout
+\begin_inset Argument 1
+status collapsed
+
+\begin_layout Plain Layout
+Strand cross-correlation plots for ChIP-seq data, before and after blacklisting.
+\end_layout
+
+\end_inset
+
 
-\series bold
 \begin_inset CommandInset label
 LatexCommand label
-name "fig:PCoA-H3K4me2-good"
+name "fig:CCF-master"
 
 \end_inset
 
-H3K4me2, SVs subtracted
+
+\series bold
+Strand cross-correlation plots for ChIP-seq data, before and after blacklisting.
 \end_layout
 
 \end_inset
@@ -4094,6 +3870,10 @@ H3K4me2, SVs subtracted
 
 \end_layout
 
+\begin_layout Standard
+\begin_inset Note Note
+status open
+
 \begin_layout Plain Layout
 \begin_inset Float figure
 wide false
@@ -4103,10 +3883,10 @@ status collapsed
 \begin_layout Plain Layout
 \align center
 \begin_inset Graphics
-	filename graphics/CD4-csaw/ChIP-seq/H3K4me3-PCA-raw-CROP.png
+	filename graphics/CD4-csaw/ChIP-seq/H3K4me2-sample-MAplot-bins-CROP.png
 	lyxscale 25
-	width 45col%
-	groupId pcoa-subfig
+	width 100col%
+	groupId colwidth-raster
 
 \end_inset
 
@@ -4121,11 +3901,11 @@ status collapsed
 \series bold
 \begin_inset CommandInset label
 LatexCommand label
-name "fig:PCoA-H3K4me3-bad"
+name "fig:MA-plot-bigbins"
 
 \end_inset
 
-H3K4me3, no correction
+MA plot of H3K4me2 read counts in 10kb bins for two arbitrary samples.
 \end_layout
 
 \end_inset
@@ -4136,137 +3916,197 @@ H3K4me3, no correction
 \end_inset
 
 
-\begin_inset space \hfill{}
+\end_layout
+
 \end_inset
 
 
-\begin_inset Float figure
-wide false
-sideways false
-status collapsed
+\end_layout
+
+\begin_layout Standard
+Promoters were defined by computing the distance from each annotated 
+\begin_inset Flex Glossary Term
+status open
 
 \begin_layout Plain Layout
-\align center
-\begin_inset Graphics
-	filename graphics/CD4-csaw/ChIP-seq/H3K4me3-PCA-SVsub-CROP.png
-	lyxscale 25
-	width 45col%
-	groupId pcoa-subfig
+TSS
+\end_layout
 
 \end_inset
 
-
-\end_layout
+ to the nearest called peak and examining the distribution of distances,
+ observing that peaks for each histone mark were enriched within a certain
+ distance of the 
+\begin_inset Flex Glossary Term
+status open
 
 \begin_layout Plain Layout
-\begin_inset Caption Standard
+TSS
+\end_layout
 
-\begin_layout Plain Layout
+\end_inset
 
-\series bold
-\begin_inset CommandInset label
-LatexCommand label
-name "fig:PCoA-H3K4me3-good"
+.
+ For H3K4me2 and H3K4me3, this distance was about 1
+\begin_inset space ~
+\end_inset
 
+kb, while for H3K27me3 it was 2.5
+\begin_inset space ~
 \end_inset
 
-H3K4me3, SVs subtracted
-\end_layout
+kb.
+ These distances were used as an 
+\begin_inset Quotes eld
+\end_inset
 
+effective promoter radius
+\begin_inset Quotes erd
 \end_inset
 
+ for each mark.
+ The promoter region for each gene was defined as the region of the genome
+ within this distance upstream or downstream of the gene's annotated 
+\begin_inset Flex Glossary Term
+status open
 
+\begin_layout Plain Layout
+TSS
 \end_layout
 
 \end_inset
 
+.
+ For genes with multiple annotated 
+\begin_inset Flex Glossary Term (pl)
+status open
 
+\begin_layout Plain Layout
+TSS
 \end_layout
 
-\begin_layout Plain Layout
-\begin_inset Float figure
-wide false
-sideways false
-status collapsed
+\end_inset
+
+, a promoter region was defined for each 
+\begin_inset Flex Glossary Term
+status open
 
 \begin_layout Plain Layout
-\align center
-\begin_inset Graphics
-	filename graphics/CD4-csaw/ChIP-seq/H3K27me3-PCA-raw-CROP.png
-	lyxscale 25
-	width 45col%
-	groupId pcoa-subfig
+TSS
+\end_layout
 
 \end_inset
 
-
-\end_layout
+ individually, and any promoters that overlapped (due to multiple 
+\begin_inset Flex Glossary Term (pl)
+status open
 
 \begin_layout Plain Layout
-\begin_inset Caption Standard
+TSS
+\end_layout
 
-\begin_layout Plain Layout
+\end_inset
 
-\series bold
-\begin_inset CommandInset label
-LatexCommand label
-name "fig:PCoA-H3K27me3-bad"
+ being closer than 2 times the radius) were merged into one large promoter.
+ Thus, some genes had multiple promoters defined, which were each analyzed
+ separately for differential modification.
+\end_layout
 
-\end_inset
+\begin_layout Standard
+Reads in promoters, peaks, and sliding windows across the genome were counted
+ and normalized using 
+\begin_inset Flex Code
+status open
 
-H3K27me3, no correction
+\begin_layout Plain Layout
+csaw
 \end_layout
 
 \end_inset
 
+ and analyzed for differential modification using 
+\begin_inset Flex Code
+status open
 
+\begin_layout Plain Layout
+edgeR
 \end_layout
 
 \end_inset
 
+ 
+\begin_inset CommandInset citation
+LatexCommand cite
+key "Lun2014,Lun2015a,Lund2012,Phipson2016"
+literal "false"
 
-\begin_inset space \hfill{}
 \end_inset
 
-
-\begin_inset Float figure
-wide false
-sideways false
-status collapsed
+.
+ Unobserved confounding factors in the 
+\begin_inset Flex Glossary Term
+status open
 
 \begin_layout Plain Layout
-\align center
-\begin_inset Graphics
-	filename graphics/CD4-csaw/ChIP-seq/H3K27me3-PCA-SVsub-CROP.png
-	lyxscale 25
-	width 45col%
-	groupId pcoa-subfig
+ChIP-seq
+\end_layout
 
 \end_inset
 
-
-\end_layout
+ data were corrected using 
+\begin_inset Flex Glossary Term
+status open
 
 \begin_layout Plain Layout
-\begin_inset Caption Standard
+SVA
+\end_layout
 
-\begin_layout Plain Layout
+\end_inset
 
-\series bold
-\begin_inset CommandInset label
-LatexCommand label
-name "fig:PCoA-H3K27me3-good"
+ 
+\begin_inset CommandInset citation
+LatexCommand cite
+key "Leek2007,Leek2014"
+literal "false"
 
 \end_inset
 
-H3K27me3, SVs subtracted
-\end_layout
+.
+ Principal coordinate plots of the promoter count data for each histone
+ mark before and after subtracting surrogate variable effects are shown
+ in Figure 
+\begin_inset CommandInset ref
+LatexCommand ref
+reference "fig:PCoA-ChIP"
+plural "false"
+caps "false"
+noprefix "false"
 
 \end_inset
 
-
+.
 \end_layout
 
+\begin_layout Standard
+\begin_inset Float figure
+wide false
+sideways false
+status collapsed
+
+\begin_layout Plain Layout
+\begin_inset Float figure
+wide false
+sideways false
+status collapsed
+
+\begin_layout Plain Layout
+\align center
+\begin_inset Graphics
+	filename graphics/CD4-csaw/ChIP-seq/H3K4me2-PCA-raw-CROP.png
+	lyxscale 25
+	width 45col%
+	groupId pcoa-subfig
+
 \end_inset
 
 
@@ -4276,27 +4116,15 @@ H3K27me3, SVs subtracted
 \begin_inset Caption Standard
 
 \begin_layout Plain Layout
-\begin_inset Argument 1
-status collapsed
-
-\begin_layout Plain Layout
-PCoA plots of ChIP-seq sliding window data, before and after subtracting
- surrogate variables (SVs).
-\end_layout
-
-\end_inset
-
 
+\series bold
 \begin_inset CommandInset label
 LatexCommand label
-name "fig:PCoA-ChIP"
+name "fig:PCoA-H3K4me2-bad"
 
 \end_inset
 
-
-\series bold
-PCoA plots of ChIP-seq sliding window data, before and after subtracting
- surrogate variables (SVs).
+H3K4me2, no correction
 \end_layout
 
 \end_inset
@@ -4307,194 +4135,135 @@ PCoA plots of ChIP-seq sliding window data, before and after subtracting
 \end_inset
 
 
-\end_layout
+\begin_inset space \hfill{}
+\end_inset
 
-\begin_layout Standard
-Reads in promoters, peaks, and sliding windows across the genome were counted
- and normalized using 
-\begin_inset Flex Code
-status open
+
+\begin_inset Float figure
+wide false
+sideways false
+status collapsed
 
 \begin_layout Plain Layout
-csaw
-\end_layout
+\align center
+\begin_inset Graphics
+	filename graphics/CD4-csaw/ChIP-seq/H3K4me2-PCA-SVsub-CROP.png
+	lyxscale 25
+	width 45col%
+	groupId pcoa-subfig
 
 \end_inset
 
- and analyzed for differential modification using 
-\begin_inset Flex Code
-status open
 
-\begin_layout Plain Layout
-edgeR
 \end_layout
 
-\end_inset
+\begin_layout Plain Layout
+\begin_inset Caption Standard
 
- 
-\begin_inset CommandInset citation
-LatexCommand cite
-key "Lun2014,Lun2015a,Lund2012,Phipson2016"
-literal "false"
+\begin_layout Plain Layout
 
-\end_inset
+\series bold
+\begin_inset CommandInset label
+LatexCommand label
+name "fig:PCoA-H3K4me2-good"
 
-.
- Unobserved confounding factors in the 
-\begin_inset Flex Glossary Term
-status open
+\end_inset
 
-\begin_layout Plain Layout
-ChIP-seq
+H3K4me2, SVs subtracted
 \end_layout
 
 \end_inset
 
- data were corrected using 
-\begin_inset Flex Glossary Term
-status open
 
-\begin_layout Plain Layout
-SVA
 \end_layout
 
 \end_inset
 
- 
-\begin_inset CommandInset citation
-LatexCommand cite
-key "Leek2007,Leek2014"
-literal "false"
-
-\end_inset
-
-.
- Principal coordinate plots of the promoter count data for each histone
- mark before and after subtracting surrogate variable effects are shown
- in Figure 
-\begin_inset CommandInset ref
-LatexCommand ref
-reference "fig:PCoA-ChIP"
-plural "false"
-caps "false"
-noprefix "false"
-
-\end_inset
 
-.
 \end_layout
 
-\begin_layout Standard
-To investigate whether the location of a peak within the promoter region
- was important, 
-\begin_inset Quotes eld
-\end_inset
-
-relative coverage profiles
-\begin_inset Quotes erd
-\end_inset
-
- were generated.
- First, 500-bp sliding windows were tiled around each annotated 
-\begin_inset Flex Glossary Term
-status open
+\begin_layout Plain Layout
+\begin_inset Float figure
+wide false
+sideways false
+status collapsed
 
 \begin_layout Plain Layout
-TSS
-\end_layout
+\align center
+\begin_inset Graphics
+	filename graphics/CD4-csaw/ChIP-seq/H3K4me3-PCA-raw-CROP.png
+	lyxscale 25
+	width 45col%
+	groupId pcoa-subfig
 
 \end_inset
 
-: one window centered on the 
-\begin_inset Flex Glossary Term
-status open
 
-\begin_layout Plain Layout
-TSS
 \end_layout
 
-\end_inset
-
- itself, and 10 windows each upstream and downstream, thus covering a 10.5-kb
- region centered on the 
-\begin_inset Flex Glossary Term
-status open
+\begin_layout Plain Layout
+\begin_inset Caption Standard
 
 \begin_layout Plain Layout
-TSS
-\end_layout
 
-\end_inset
+\series bold
+\begin_inset CommandInset label
+LatexCommand label
+name "fig:PCoA-H3K4me3-bad"
 
- with 21 windows.
- Reads in each window for each 
-\begin_inset Flex Glossary Term
-status open
+\end_inset
 
-\begin_layout Plain Layout
-TSS
+H3K4me3, no correction
 \end_layout
 
 \end_inset
 
- were counted in each sample, and the counts were normalized and converted
- to 
-\begin_inset Flex Glossary Term
-status open
 
-\begin_layout Plain Layout
-logCPM
 \end_layout
 
 \end_inset
 
- as in the differential modification analysis.
- Then, the 
-\begin_inset Flex Glossary Term
-status open
-
-\begin_layout Plain Layout
-logCPM
-\end_layout
 
+\begin_inset space \hfill{}
 \end_inset
 
- values within each promoter were normalized to an average of zero, such
- that each window's normalized abundance now represents the relative read
- depth of that window compared to all other windows in the same promoter.
- The normalized abundance values for each window in a promoter are collectively
- referred to as that promoter's 
-\begin_inset Quotes eld
-\end_inset
 
-relative coverage profile
-\begin_inset Quotes erd
+\begin_inset Float figure
+wide false
+sideways false
+status collapsed
+
+\begin_layout Plain Layout
+\align center
+\begin_inset Graphics
+	filename graphics/CD4-csaw/ChIP-seq/H3K4me3-PCA-SVsub-CROP.png
+	lyxscale 25
+	width 45col%
+	groupId pcoa-subfig
+
 \end_inset
 
-.
-\end_layout
 
-\begin_layout Subsection
-MOFA recovers biologically relevant variation from blind analysis by correlating
- across datasets
 \end_layout
 
-\begin_layout Standard
-\begin_inset ERT
-status open
+\begin_layout Plain Layout
+\begin_inset Caption Standard
 
 \begin_layout Plain Layout
 
+\series bold
+\begin_inset CommandInset label
+LatexCommand label
+name "fig:PCoA-H3K4me3-good"
 
-\backslash
-afterpage{
+\end_inset
+
+H3K4me3, SVs subtracted
 \end_layout
 
-\begin_layout Plain Layout
+\end_inset
 
 
-\backslash
-begin{landscape}
 \end_layout
 
 \end_inset
@@ -4502,25 +4271,19 @@ begin{landscape}
 
 \end_layout
 
-\begin_layout Standard
-\begin_inset Float figure
-wide false
-sideways false
-status open
-
 \begin_layout Plain Layout
 \begin_inset Float figure
 wide false
 sideways false
-status open
+status collapsed
 
 \begin_layout Plain Layout
 \align center
 \begin_inset Graphics
-	filename graphics/CD4-csaw/MOFA-varExplaiend-matrix-CROP.png
+	filename graphics/CD4-csaw/ChIP-seq/H3K27me3-PCA-raw-CROP.png
 	lyxscale 25
 	width 45col%
-	groupId mofa-subfig
+	groupId pcoa-subfig
 
 \end_inset
 
@@ -4535,25 +4298,11 @@ status open
 \series bold
 \begin_inset CommandInset label
 LatexCommand label
-name "fig:mofa-varexplained"
+name "fig:PCoA-H3K27me3-bad"
 
 \end_inset
 
-Variance explained in each data set by each latent factor estimated by MOFA.
-
-\series default
- For each LF learned by MOFA, the variance explained by that factor in each
- data set (
-\begin_inset Quotes eld
-\end_inset
-
-view
-\begin_inset Quotes erd
-\end_inset
-
-) is shown by the shading of the cells in the lower section.
- The upper section shows the total fraction of each data set's variance
- that is explained by all LFs combined.
+H3K27me3, no correction
 \end_layout
 
 \end_inset
@@ -4571,15 +4320,15 @@ view
 \begin_inset Float figure
 wide false
 sideways false
-status open
+status collapsed
 
 \begin_layout Plain Layout
 \align center
 \begin_inset Graphics
-	filename graphics/CD4-csaw/MOFA-LF-scatter-CROP.png
+	filename graphics/CD4-csaw/ChIP-seq/H3K27me3-PCA-SVsub-CROP.png
 	lyxscale 25
 	width 45col%
-	groupId mofa-subfig
+	groupId pcoa-subfig
 
 \end_inset
 
@@ -4594,16 +4343,11 @@ status open
 \series bold
 \begin_inset CommandInset label
 LatexCommand label
-name "fig:mofa-lf-scatter"
+name "fig:PCoA-H3K27me3-good"
 
 \end_inset
 
-Scatter plots of specific pairs of MOFA latent factors.
-
-\series default
- LFs 1, 4, and 5 explain substantial variation in all data sets, so they
- are plotted against each other in order to reveal patterns of variation
- that are shared across all data sets.
+H3K27me3, SVs subtracted
 \end_layout
 
 \end_inset
@@ -4624,7 +4368,8 @@ Scatter plots of specific pairs of MOFA latent factors.
 status collapsed
 
 \begin_layout Plain Layout
-MOFA latent factors identify shared patterns of variation.
+PCoA plots of ChIP-seq sliding window data, before and after subtracting
+ surrogate variables (SVs).
 \end_layout
 
 \end_inset
@@ -4632,13 +4377,14 @@ MOFA latent factors identify shared patterns of variation.
 
 \begin_inset CommandInset label
 LatexCommand label
-name "fig:MOFA-master"
+name "fig:PCoA-ChIP"
 
 \end_inset
 
 
 \series bold
-MOFA latent factors identify shared patterns of variation.
+PCoA plots of ChIP-seq sliding window data, before and after subtracting
+ surrogate variables (SVs).
 \end_layout
 
 \end_inset
@@ -4652,24 +4398,98 @@ MOFA latent factors identify shared patterns of variation.
 \end_layout
 
 \begin_layout Standard
-\begin_inset ERT
+To investigate whether the location of a peak within the promoter region
+ was important, 
+\begin_inset Quotes eld
+\end_inset
+
+relative coverage profiles
+\begin_inset Quotes erd
+\end_inset
+
+ were generated.
+ First, 500-bp sliding windows were tiled around each annotated 
+\begin_inset Flex Glossary Term
 status open
 
 \begin_layout Plain Layout
+TSS
+\end_layout
 
+\end_inset
 
-\backslash
-end{landscape}
+: one window centered on the 
+\begin_inset Flex Glossary Term
+status open
+
+\begin_layout Plain Layout
+TSS
+\end_layout
+
+\end_inset
+
+ itself, and 10 windows each upstream and downstream, thus covering a 10.5-kb
+ region centered on the 
+\begin_inset Flex Glossary Term
+status open
+
+\begin_layout Plain Layout
+TSS
+\end_layout
+
+\end_inset
+
+ with 21 windows.
+ Reads in each window for each 
+\begin_inset Flex Glossary Term
+status open
+
+\begin_layout Plain Layout
+TSS
 \end_layout
 
+\end_inset
+
+ were counted in each sample, and the counts were normalized and converted
+ to 
+\begin_inset Flex Glossary Term
+status open
+
 \begin_layout Plain Layout
+logCPM
+\end_layout
 
-}
+\end_inset
+
+ as in the differential modification analysis.
+ Then, the 
+\begin_inset Flex Glossary Term
+status open
+
+\begin_layout Plain Layout
+logCPM
 \end_layout
 
 \end_inset
 
+ values within each promoter were normalized to an average of zero, such
+ that each window's normalized abundance now represents the relative read
+ depth of that window compared to all other windows in the same promoter.
+ The normalized abundance values for each window in a promoter are collectively
+ referred to as that promoter's 
+\begin_inset Quotes eld
+\end_inset
+
+relative coverage profile
+\begin_inset Quotes erd
+\end_inset
+
+.
+\end_layout
 
+\begin_layout Subsection
+MOFA recovers biologically relevant variation from blind analysis by correlating
+ across datasets
 \end_layout
 
 \begin_layout Standard
@@ -4816,7 +4636,32 @@ noprefix "false"
 \end_layout
 
 \begin_layout Standard
-\begin_inset Note Note
+\begin_inset ERT
+status open
+
+\begin_layout Plain Layout
+
+
+\backslash
+afterpage{
+\end_layout
+
+\begin_layout Plain Layout
+
+
+\backslash
+begin{landscape}
+\end_layout
+
+\end_inset
+
+
+\end_layout
+
+\begin_layout Standard
+\begin_inset Float figure
+wide false
+sideways false
 status collapsed
 
 \begin_layout Plain Layout
@@ -4828,10 +4673,10 @@ status open
 \begin_layout Plain Layout
 \align center
 \begin_inset Graphics
-	filename graphics/CD4-csaw/MOFA-batch-correct-CROP.png
+	filename graphics/CD4-csaw/MOFA-varExplaiend-matrix-CROP.png
 	lyxscale 25
-	width 100col%
-	groupId colwidth-raster
+	width 45col%
+	groupId mofa-subfig
 
 \end_inset
 
@@ -4846,16 +4691,25 @@ status open
 \series bold
 \begin_inset CommandInset label
 LatexCommand label
-name "fig:mofa-batchsub"
+name "fig:mofa-varexplained"
 
 \end_inset
 
-Result of RNA-seq batch-correction using MOFA latent factors
-\end_layout
+Variance explained in each data set by each latent factor estimated by MOFA.
 
+\series default
+ For each LF learned by MOFA, the variance explained by that factor in each
+ data set (
+\begin_inset Quotes eld
 \end_inset
 
+view
+\begin_inset Quotes erd
+\end_inset
 
+) is shown by the shading of the cells in the lower section.
+ The upper section shows the total fraction of each data set's variance
+ that is explained by all LFs combined.
 \end_layout
 
 \end_inset
@@ -4866,121 +4720,334 @@ Result of RNA-seq batch-correction using MOFA latent factors
 \end_inset
 
 
-\end_layout
+\begin_inset space \hfill{}
+\end_inset
 
-\begin_layout Standard
-\begin_inset Note Note
+
+\begin_inset Float figure
+wide false
+sideways false
 status open
 
 \begin_layout Plain Layout
-Placing these floats is a challenge
-\end_layout
+\align center
+\begin_inset Graphics
+	filename graphics/CD4-csaw/MOFA-LF-scatter-CROP.png
+	lyxscale 25
+	width 45col%
+	groupId mofa-subfig
 
 \end_inset
 
 
 \end_layout
 
-\begin_layout Standard
-\begin_inset Float table
-wide false
-sideways false
-status collapsed
-
 \begin_layout Plain Layout
-\align center
-\begin_inset Tabular
-<lyxtabular version="3" rows="11" columns="3">
-<features tabularvalignment="middle">
-<column alignment="center" valignment="top">
-<column alignment="center" valignment="top">
-<column alignment="center" valignment="top">
-<row>
-<cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
-\begin_inset Text
+\begin_inset Caption Standard
 
 \begin_layout Plain Layout
-Test
-\end_layout
+
+\series bold
+\begin_inset CommandInset label
+LatexCommand label
+name "fig:mofa-lf-scatter"
 
 \end_inset
-</cell>
-<cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
-\begin_inset Text
 
-\begin_layout Plain Layout
-Est.
- non-null
-\end_layout
+Scatter plots of specific pairs of MOFA latent factors.
 
-\end_inset
-</cell>
-<cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" rightline="true" usebox="none">
-\begin_inset Text
+\series default
+ LFs 1, 4, and 5 explain substantial variation in all data sets, so they
+ are plotted against each other in order to reveal patterns of variation
+ that are shared across all data sets.
+\end_layout
 
-\begin_layout Plain Layout
-\begin_inset Formula $\mathrm{FDR}\le10\%$
 \end_inset
 
 
 \end_layout
 
 \end_inset
-</cell>
-</row>
-<row>
-<cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
-\begin_inset Text
 
-\begin_layout Plain Layout
-Naïve Day 0 vs Day 1
-\end_layout
 
-\end_inset
-</cell>
-<cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
-\begin_inset Text
+\end_layout
 
 \begin_layout Plain Layout
-5992
-\end_layout
+\begin_inset Caption Standard
 
-\end_inset
-</cell>
-<cell alignment="center" valignment="top" topline="true" leftline="true" rightline="true" usebox="none">
-\begin_inset Text
+\begin_layout Plain Layout
+\begin_inset Argument 1
+status collapsed
 
 \begin_layout Plain Layout
-1613
+MOFA latent factors identify shared patterns of variation.
 \end_layout
 
 \end_inset
-</cell>
-</row>
-<row>
-<cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
+
+
+\begin_inset CommandInset label
+LatexCommand label
+name "fig:MOFA-master"
+
+\end_inset
+
+
+\series bold
+MOFA latent factors identify shared patterns of variation.
+\end_layout
+
+\end_inset
+
+
+\end_layout
+
+\end_inset
+
+
+\end_layout
+
+\begin_layout Standard
+\begin_inset ERT
+status open
+
+\begin_layout Plain Layout
+
+
+\backslash
+end{landscape}
+\end_layout
+
+\begin_layout Plain Layout
+
+}
+\end_layout
+
+\end_inset
+
+
+\end_layout
+
+\begin_layout Standard
+\begin_inset Note Note
+status collapsed
+
+\begin_layout Plain Layout
+\begin_inset Float figure
+wide false
+sideways false
+status open
+
+\begin_layout Plain Layout
+\align center
+\begin_inset Graphics
+	filename graphics/CD4-csaw/MOFA-batch-correct-CROP.png
+	lyxscale 25
+	width 100col%
+	groupId colwidth-raster
+
+\end_inset
+
+
+\end_layout
+
+\begin_layout Plain Layout
+\begin_inset Caption Standard
+
+\begin_layout Plain Layout
+
+\series bold
+\begin_inset CommandInset label
+LatexCommand label
+name "fig:mofa-batchsub"
+
+\end_inset
+
+Result of RNA-seq batch-correction using MOFA latent factors
+\end_layout
+
+\end_inset
+
+
+\end_layout
+
+\end_inset
+
+
+\end_layout
+
+\end_inset
+
+
+\end_layout
+
+\begin_layout Section
+Results
+\end_layout
+
+\begin_layout Standard
+\begin_inset Flex TODO Note (inline)
+status open
+
+\begin_layout Plain Layout
+Focus on what hypotheses were tested, then select figures that show how
+ those hypotheses were tested, even if the result is a negative.
+ Not every interesting result needs to be in here.
+ Chapter should tell a story.
+ 
+\end_layout
+
+\end_inset
+
+
+\end_layout
+
+\begin_layout Subsection
+Interpretation of RNA-seq analysis is limited by a major confounding factor
+\end_layout
+
+\begin_layout Standard
+\begin_inset Note Note
+status open
+
+\begin_layout Plain Layout
+Putting a float here causes an error.
+ No idea why.
+ See above for the floats that should be placed here.
+\end_layout
+
+\end_inset
+
+
+\end_layout
+
+\begin_layout Standard
+Genes called as present in the 
+\begin_inset Flex Glossary Term
+status open
+
+\begin_layout Plain Layout
+RNA-seq
+\end_layout
+
+\end_inset
+
+ data were tested for differential expression between all time points and
+ cell types.
+ The counts of differentially expressed genes are shown in Table 
+\begin_inset CommandInset ref
+LatexCommand ref
+reference "tab:Estimated-and-detected-rnaseq"
+plural "false"
+caps "false"
+noprefix "false"
+
+\end_inset
+
+.
+ Notably, all the results for Day 0 and Day 5 have substantially fewer genes
+ called differentially expressed than any of the results for other time
+ points.
+ This is an unfortunate result of the difference in sample quality between
+ the two batches of 
+\begin_inset Flex Glossary Term
+status open
+
+\begin_layout Plain Layout
+RNA-seq
+\end_layout
+
+\end_inset
+
+ data.
+ All the samples in Batch 1, which includes all the samples from Days 0
+ and 5, have substantially more variability than the samples in Batch 2,
+ which includes the other time points.
+ This is reflected in the substantially higher weights assigned to Batch
+ 2 (Figure 
+\begin_inset CommandInset ref
+LatexCommand ref
+reference "fig:RNA-seq-weights-vs-covars"
+plural "false"
+caps "false"
+noprefix "false"
+
+\end_inset
+
+).
+ The batch effect has both a systematic component and a random noise component.
+ While the systematic component was subtracted out using ComBat (Figure
+ 
+\begin_inset CommandInset ref
+LatexCommand ref
+reference "fig:RNA-PCA"
+plural "false"
+caps "false"
+noprefix "false"
+
+\end_inset
+
+), no such correction is possible for the noise component: Batch 1 simply
+ has substantially more random noise in it, which reduces the statistical
+ power for any differential expression tests involving samples in that batch.
+ 
+\end_layout
+
+\begin_layout Standard
+\begin_inset Note Note
+status open
+
+\begin_layout Plain Layout
+Placing these floats is a challenge
+\end_layout
+
+\end_inset
+
+
+\end_layout
+
+\begin_layout Standard
+\begin_inset Float table
+wide false
+sideways false
+status collapsed
+
+\begin_layout Plain Layout
+\align center
+\begin_inset Tabular
+<lyxtabular version="3" rows="11" columns="3">
+<features tabularvalignment="middle">
+<column alignment="center" valignment="top">
+<column alignment="center" valignment="top">
+<column alignment="center" valignment="top">
+<row>
+<cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
 \begin_inset Text
 
 \begin_layout Plain Layout
-Naïve Day 0 vs Day 5
+Test
 \end_layout
 
 \end_inset
 </cell>
-<cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
+<cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
 \begin_inset Text
 
 \begin_layout Plain Layout
-3038
+Est.
+ non-null
 \end_layout
 
 \end_inset
 </cell>
-<cell alignment="center" valignment="top" topline="true" leftline="true" rightline="true" usebox="none">
+<cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" rightline="true" usebox="none">
 \begin_inset Text
 
 \begin_layout Plain Layout
-32
+\begin_inset Formula $\mathrm{FDR}\le10\%$
+\end_inset
+
+
 \end_layout
 
 \end_inset
@@ -4991,83 +5058,607 @@ Naïve Day 0 vs Day 5
 \begin_inset Text
 
 \begin_layout Plain Layout
-Naïve Day 0 vs Day 14
+Naïve Day 0 vs Day 1
+\end_layout
+
+\end_inset
+</cell>
+<cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
+\begin_inset Text
+
+\begin_layout Plain Layout
+5992
+\end_layout
+
+\end_inset
+</cell>
+<cell alignment="center" valignment="top" topline="true" leftline="true" rightline="true" usebox="none">
+\begin_inset Text
+
+\begin_layout Plain Layout
+1613
+\end_layout
+
+\end_inset
+</cell>
+</row>
+<row>
+<cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
+\begin_inset Text
+
+\begin_layout Plain Layout
+Naïve Day 0 vs Day 5
+\end_layout
+
+\end_inset
+</cell>
+<cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
+\begin_inset Text
+
+\begin_layout Plain Layout
+3038
+\end_layout
+
+\end_inset
+</cell>
+<cell alignment="center" valignment="top" topline="true" leftline="true" rightline="true" usebox="none">
+\begin_inset Text
+
+\begin_layout Plain Layout
+32
+\end_layout
+
+\end_inset
+</cell>
+</row>
+<row>
+<cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
+\begin_inset Text
+
+\begin_layout Plain Layout
+Naïve Day 0 vs Day 14
+\end_layout
+
+\end_inset
+</cell>
+<cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
+\begin_inset Text
+
+\begin_layout Plain Layout
+1870
+\end_layout
+
+\end_inset
+</cell>
+<cell alignment="center" valignment="top" topline="true" leftline="true" rightline="true" usebox="none">
+\begin_inset Text
+
+\begin_layout Plain Layout
+190
+\end_layout
+
+\end_inset
+</cell>
+</row>
+<row>
+<cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
+\begin_inset Text
+
+\begin_layout Plain Layout
+Memory Day 0 vs Day 1
+\end_layout
+
+\end_inset
+</cell>
+<cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
+\begin_inset Text
+
+\begin_layout Plain Layout
+3195
+\end_layout
+
+\end_inset
+</cell>
+<cell alignment="center" valignment="top" topline="true" leftline="true" rightline="true" usebox="none">
+\begin_inset Text
+
+\begin_layout Plain Layout
+411
+\end_layout
+
+\end_inset
+</cell>
+</row>
+<row>
+<cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
+\begin_inset Text
+
+\begin_layout Plain Layout
+Memory Day 0 vs Day 5
+\end_layout
+
+\end_inset
+</cell>
+<cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
+\begin_inset Text
+
+\begin_layout Plain Layout
+2688
+\end_layout
+
+\end_inset
+</cell>
+<cell alignment="center" valignment="top" topline="true" leftline="true" rightline="true" usebox="none">
+\begin_inset Text
+
+\begin_layout Plain Layout
+18
+\end_layout
+
+\end_inset
+</cell>
+</row>
+<row>
+<cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
+\begin_inset Text
+
+\begin_layout Plain Layout
+Memory Day 0 vs Day 14
+\end_layout
+
+\end_inset
+</cell>
+<cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
+\begin_inset Text
+
+\begin_layout Plain Layout
+1911
+\end_layout
+
+\end_inset
+</cell>
+<cell alignment="center" valignment="top" topline="true" leftline="true" rightline="true" usebox="none">
+\begin_inset Text
+
+\begin_layout Plain Layout
+227
+\end_layout
+
+\end_inset
+</cell>
+</row>
+<row>
+<cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
+\begin_inset Text
+
+\begin_layout Plain Layout
+Day 0 Naïve vs Memory
+\end_layout
+
+\end_inset
+</cell>
+<cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
+\begin_inset Text
+
+\begin_layout Plain Layout
+0
+\end_layout
+
+\end_inset
+</cell>
+<cell alignment="center" valignment="top" topline="true" leftline="true" rightline="true" usebox="none">
+\begin_inset Text
+
+\begin_layout Plain Layout
+2
+\end_layout
+
+\end_inset
+</cell>
+</row>
+<row>
+<cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
+\begin_inset Text
+
+\begin_layout Plain Layout
+Day 1 Naïve vs Memory
+\end_layout
+
+\end_inset
+</cell>
+<cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
+\begin_inset Text
+
+\begin_layout Plain Layout
+9167
+\end_layout
+
+\end_inset
+</cell>
+<cell alignment="center" valignment="top" topline="true" leftline="true" rightline="true" usebox="none">
+\begin_inset Text
+
+\begin_layout Plain Layout
+5532
+\end_layout
+
+\end_inset
+</cell>
+</row>
+<row>
+<cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
+\begin_inset Text
+
+\begin_layout Plain Layout
+Day 5 Naïve vs Memory
+\end_layout
+
+\end_inset
+</cell>
+<cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
+\begin_inset Text
+
+\begin_layout Plain Layout
+0
+\end_layout
+
+\end_inset
+</cell>
+<cell alignment="center" valignment="top" topline="true" leftline="true" rightline="true" usebox="none">
+\begin_inset Text
+
+\begin_layout Plain Layout
+0
+\end_layout
+
+\end_inset
+</cell>
+</row>
+<row>
+<cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
+\begin_inset Text
+
+\begin_layout Plain Layout
+Day 14 Naïve vs Memory
+\end_layout
+
+\end_inset
+</cell>
+<cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
+\begin_inset Text
+
+\begin_layout Plain Layout
+6446
+\end_layout
+
+\end_inset
+</cell>
+<cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" rightline="true" usebox="none">
+\begin_inset Text
+
+\begin_layout Plain Layout
+2319
+\end_layout
+
+\end_inset
+</cell>
+</row>
+</lyxtabular>
+
+\end_inset
+
+
+\end_layout
+
+\begin_layout Plain Layout
+\begin_inset Caption Standard
+
+\begin_layout Plain Layout
+\begin_inset Argument 1
+status collapsed
+
+\begin_layout Plain Layout
+Estimated and detected differentially expressed genes.
+\end_layout
+
+\end_inset
+
+
+\begin_inset CommandInset label
+LatexCommand label
+name "tab:Estimated-and-detected-rnaseq"
+
+\end_inset
+
+
+\series bold
+Estimated and detected differentially expressed genes.
+
+\series default
+ 
+\begin_inset Quotes eld
+\end_inset
+
+Test
+\begin_inset Quotes erd
+\end_inset
+
+: Which sample groups were compared; 
+\begin_inset Quotes eld
+\end_inset
+
+Est non-null
+\begin_inset Quotes erd
+\end_inset
+
+: Estimated number of differentially expressed genes, using the method of
+ averaging local FDR values 
+\begin_inset CommandInset citation
+LatexCommand cite
+key "Phipson2013Thesis"
+literal "false"
+
+\end_inset
+
+; 
+\begin_inset Quotes eld
+\end_inset
+
+
+\begin_inset Formula $\mathrm{FDR}\le10\%$
+\end_inset
+
+
+\begin_inset Quotes erd
+\end_inset
+
+: Number of significantly differentially expressed genes at an FDR threshold
+ of 10%.
+ The total number of genes tested was 16707.
+\end_layout
+
+\end_inset
+
+
+\end_layout
+
+\end_inset
+
+
+\end_layout
+
+\begin_layout Standard
+Despite the difficulty in detecting specific differentially expressed genes,
+ there is still evidence that differential expression is present for these
+ time points.
+ In Figure 
+\begin_inset CommandInset ref
+LatexCommand ref
+reference "fig:rna-pca-final"
+plural "false"
+caps "false"
+noprefix "false"
+
+\end_inset
+
+, there is a clear separation between naïve and memory samples at Day 0,
+ despite the fact that only 2 genes were significantly differentially expressed
+ for this comparison.
+ Similarly, the small numbers of genes detected for the Day 0 vs Day 5 compariso
+ns do not reflect the large separation between these time points in Figure
+ 
+\begin_inset CommandInset ref
+LatexCommand ref
+reference "fig:rna-pca-final"
+plural "false"
+caps "false"
+noprefix "false"
+
+\end_inset
+
+.
+ In addition, the 
+\begin_inset Flex Glossary Term
+status open
+
+\begin_layout Plain Layout
+MOFA
+\end_layout
+
+\end_inset
+
+ 
+\begin_inset Flex Glossary Term
+status open
+
+\begin_layout Plain Layout
+LF
+\end_layout
+
+\end_inset
+
+ plots in Figure 
+\begin_inset CommandInset ref
+LatexCommand ref
+reference "fig:mofa-lf-scatter"
+plural "false"
+caps "false"
+noprefix "false"
+
+\end_inset
+
+.
+ This suggests that there is indeed a differential expression signal present
+ in the data for these comparisons, but the large variability in the Batch
+ 1 samples obfuscates this signal at the individual gene level.
+ As a result, it is impossible to make any meaningful statements about the
+ 
+\begin_inset Quotes eld
+\end_inset
+
+size
+\begin_inset Quotes erd
+\end_inset
+
+ of the gene signature for any time point, since the number of significant
+ genes as well as the estimated number of differentially expressed genes
+ depends so strongly on the variations in sample quality in addition to
+ the size of the differential expression signal in the data.
+ Gene-set enrichment analyses are similarly impractical.
+ However, analyses looking at genome-wide patterns of expression are still
+ practical.
+\end_layout
+
+\begin_layout Standard
+\begin_inset Float figure
+wide false
+sideways false
+status collapsed
+
+\begin_layout Plain Layout
+\align center
+\begin_inset Graphics
+	filename graphics/CD4-csaw/RNA-seq/PCA-final-12-CROP.png
+	lyxscale 25
+	width 100col%
+	groupId colwidth-raster
+
+\end_inset
+
+
+\end_layout
+
+\begin_layout Plain Layout
+\begin_inset Caption Standard
+
+\begin_layout Plain Layout
+\begin_inset Argument 1
+status collapsed
+
+\begin_layout Plain Layout
+PCoA plot of RNA-seq samples after ComBat batch correction.
+\end_layout
+
+\end_inset
+
+
+\begin_inset CommandInset label
+LatexCommand label
+name "fig:rna-pca-final"
+
+\end_inset
+
+
+\series bold
+PCoA plot of RNA-seq samples after ComBat batch correction.
+ 
+\series default
+Each point represents an individual sample.
+ Samples with the same combination of cell type and time point are encircled
+ with a shaded region to aid in visual identification of the sample groups.
+ Samples with of same cell type from the same donor are connected by lines
+ to indicate the 
+\begin_inset Quotes eld
+\end_inset
+
+trajectory
+\begin_inset Quotes erd
+\end_inset
+
+ of each donor's cells over time in PCoA space.
+\end_layout
+
+\end_inset
+
+
+\end_layout
+
+\end_inset
+
+
+\end_layout
+
+\begin_layout Subsection
+H3K4 and H3K27 methylation occur in broad regions and are enriched near
+ promoters
 \end_layout
 
-\end_inset
-</cell>
-<cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
-\begin_inset Text
+\begin_layout Standard
+\begin_inset Float table
+wide false
+sideways false
+status open
 
 \begin_layout Plain Layout
-1870
-\end_layout
-
-\end_inset
-</cell>
-<cell alignment="center" valignment="top" topline="true" leftline="true" rightline="true" usebox="none">
-\begin_inset Text
+\align center
+\begin_inset Flex TODO Note (inline)
+status open
 
 \begin_layout Plain Layout
-190
+Also get 
+\emph on
+median
+\emph default
+ peak width and maybe other quantiles (25%, 75%)
 \end_layout
 
 \end_inset
-</cell>
-</row>
-<row>
-<cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
-\begin_inset Text
 
-\begin_layout Plain Layout
-Memory Day 0 vs Day 1
+
 \end_layout
 
-\end_inset
-</cell>
-<cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
+\begin_layout Plain Layout
+\align center
+\begin_inset Tabular
+<lyxtabular version="3" rows="4" columns="5">
+<features tabularvalignment="middle">
+<column alignment="center" valignment="top">
+<column alignment="center" valignment="top">
+<column alignment="center" valignment="top">
+<column alignment="center" valignment="top">
+<column alignment="center" valignment="top">
+<row>
+<cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
 \begin_inset Text
 
 \begin_layout Plain Layout
-3195
+Histone Mark
 \end_layout
 
 \end_inset
 </cell>
-<cell alignment="center" valignment="top" topline="true" leftline="true" rightline="true" usebox="none">
+<cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
 \begin_inset Text
 
 \begin_layout Plain Layout
-411
+# Peaks
 \end_layout
 
 \end_inset
 </cell>
-</row>
-<row>
-<cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
+<cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
 \begin_inset Text
 
 \begin_layout Plain Layout
-Memory Day 0 vs Day 5
+Mean peak width
 \end_layout
 
 \end_inset
 </cell>
-<cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
+<cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
 \begin_inset Text
 
 \begin_layout Plain Layout
-2688
+genome coverage
 \end_layout
 
 \end_inset
 </cell>
-<cell alignment="center" valignment="top" topline="true" leftline="true" rightline="true" usebox="none">
+<cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" rightline="true" usebox="none">
 \begin_inset Text
 
 \begin_layout Plain Layout
-18
+FRiP
 \end_layout
 
 \end_inset
@@ -5078,7 +5669,7 @@ Memory Day 0 vs Day 5
 \begin_inset Text
 
 \begin_layout Plain Layout
-Memory Day 0 vs Day 14
+H3K4me2
 \end_layout
 
 \end_inset
@@ -5087,27 +5678,16 @@ Memory Day 0 vs Day 14
 \begin_inset Text
 
 \begin_layout Plain Layout
-1911
-\end_layout
-
-\end_inset
-</cell>
-<cell alignment="center" valignment="top" topline="true" leftline="true" rightline="true" usebox="none">
-\begin_inset Text
-
-\begin_layout Plain Layout
-227
+14965
 \end_layout
 
 \end_inset
 </cell>
-</row>
-<row>
 <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
 \begin_inset Text
 
 \begin_layout Plain Layout
-Day 0 Naïve vs Memory
+3970
 \end_layout
 
 \end_inset
@@ -5116,7 +5696,7 @@ Day 0 Naïve vs Memory
 \begin_inset Text
 
 \begin_layout Plain Layout
-0
+1.92%
 \end_layout
 
 \end_inset
@@ -5125,7 +5705,7 @@ Day 0 Naïve vs Memory
 \begin_inset Text
 
 \begin_layout Plain Layout
-2
+14.2%
 \end_layout
 
 \end_inset
@@ -5136,7 +5716,7 @@ Day 0 Naïve vs Memory
 \begin_inset Text
 
 \begin_layout Plain Layout
-Day 1 Naïve vs Memory
+H3K4me3
 \end_layout
 
 \end_inset
@@ -5145,27 +5725,16 @@ Day 1 Naïve vs Memory
 \begin_inset Text
 
 \begin_layout Plain Layout
-9167
-\end_layout
-
-\end_inset
-</cell>
-<cell alignment="center" valignment="top" topline="true" leftline="true" rightline="true" usebox="none">
-\begin_inset Text
-
-\begin_layout Plain Layout
-5532
+6163
 \end_layout
 
 \end_inset
 </cell>
-</row>
-<row>
 <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
 \begin_inset Text
 
 \begin_layout Plain Layout
-Day 5 Naïve vs Memory
+2946
 \end_layout
 
 \end_inset
@@ -5174,7 +5743,7 @@ Day 5 Naïve vs Memory
 \begin_inset Text
 
 \begin_layout Plain Layout
-0
+0.588%
 \end_layout
 
 \end_inset
@@ -5183,7 +5752,7 @@ Day 5 Naïve vs Memory
 \begin_inset Text
 
 \begin_layout Plain Layout
-0
+6.57%
 \end_layout
 
 \end_inset
@@ -5194,7 +5763,7 @@ Day 5 Naïve vs Memory
 \begin_inset Text
 
 \begin_layout Plain Layout
-Day 14 Naïve vs Memory
+H3K27me3
 \end_layout
 
 \end_inset
@@ -5203,230 +5772,53 @@ Day 14 Naïve vs Memory
 \begin_inset Text
 
 \begin_layout Plain Layout
-6446
-\end_layout
-
-\end_inset
-</cell>
-<cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" rightline="true" usebox="none">
-\begin_inset Text
-
-\begin_layout Plain Layout
-2319
+18139
 \end_layout
 
 \end_inset
 </cell>
-</row>
-</lyxtabular>
-
-\end_inset
-
-
-\end_layout
-
-\begin_layout Plain Layout
-\begin_inset Caption Standard
-
-\begin_layout Plain Layout
-\begin_inset Argument 1
-status collapsed
-
-\begin_layout Plain Layout
-Estimated and detected differentially expressed genes.
-\end_layout
-
-\end_inset
-
-
-\begin_inset CommandInset label
-LatexCommand label
-name "tab:Estimated-and-detected-rnaseq"
-
-\end_inset
-
-
-\series bold
-Estimated and detected differentially expressed genes.
-
-\series default
- 
-\begin_inset Quotes eld
-\end_inset
-
-Test
-\begin_inset Quotes erd
-\end_inset
-
-: Which sample groups were compared; 
-\begin_inset Quotes eld
-\end_inset
-
-Est non-null
-\begin_inset Quotes erd
-\end_inset
-
-: Estimated number of differentially expressed genes, using the method of
- averaging local FDR values 
-\begin_inset CommandInset citation
-LatexCommand cite
-key "Phipson2013Thesis"
-literal "false"
-
-\end_inset
-
-; 
-\begin_inset Quotes eld
-\end_inset
-
-
-\begin_inset Formula $\mathrm{FDR}\le10\%$
-\end_inset
-
-
-\begin_inset Quotes erd
-\end_inset
-
-: Number of significantly differentially expressed genes at an FDR threshold
- of 10%.
- The total number of genes tested was 16707.
-\end_layout
-
-\end_inset
-
-
-\end_layout
-
-\end_inset
-
-
-\end_layout
-
-\begin_layout Section
-Results
-\end_layout
-
-\begin_layout Standard
-\begin_inset Flex TODO Note (inline)
-status open
-
-\begin_layout Plain Layout
-Focus on what hypotheses were tested, then select figures that show how
- those hypotheses were tested, even if the result is a negative.
- Not every interesting result needs to be in here.
- Chapter should tell a story.
- 
-\end_layout
-
-\end_inset
-
-
-\end_layout
-
-\begin_layout Subsection
-Interpretation of RNA-seq analysis is limited by a major confounding factor
-\end_layout
-
-\begin_layout Standard
-\begin_inset Note Note
-status open
-
-\begin_layout Plain Layout
-Putting a float here causes an error.
- No idea why.
- See above for the floats that should be placed here.
-\end_layout
-
-\end_inset
-
-
-\end_layout
-
-\begin_layout Standard
-Genes called as present in the 
-\begin_inset Flex Glossary Term
-status open
+<cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
+\begin_inset Text
 
 \begin_layout Plain Layout
-RNA-seq
+18967
 \end_layout
 
 \end_inset
+</cell>
+<cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
+\begin_inset Text
 
- data were tested for differential expression between all time points and
- cell types.
- The counts of differentially expressed genes are shown in Table 
-\begin_inset CommandInset ref
-LatexCommand ref
-reference "tab:Estimated-and-detected-rnaseq"
-plural "false"
-caps "false"
-noprefix "false"
+\begin_layout Plain Layout
+11.1%
+\end_layout
 
 \end_inset
-
-.
- Notably, all the results for Day 0 and Day 5 have substantially fewer genes
- called differentially expressed than any of the results for other time
- points.
- This is an unfortunate result of the difference in sample quality between
- the two batches of 
-\begin_inset Flex Glossary Term
-status open
+</cell>
+<cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" rightline="true" usebox="none">
+\begin_inset Text
 
 \begin_layout Plain Layout
-RNA-seq
+22.5%
 \end_layout
 
 \end_inset
-
- data.
- All the samples in Batch 1, which includes all the samples from Days 0
- and 5, have substantially more variability than the samples in Batch 2,
- which includes the other time points.
- This is reflected in the substantially higher weights assigned to Batch
- 2 (Figure 
-\begin_inset CommandInset ref
-LatexCommand ref
-reference "fig:RNA-seq-weights-vs-covars"
-plural "false"
-caps "false"
-noprefix "false"
+</cell>
+</row>
+</lyxtabular>
 
 \end_inset
 
-).
- The batch effect has both a systematic component and a random noise component.
- While the systematic component was subtracted out using ComBat (Figure
- 
-\begin_inset CommandInset ref
-LatexCommand ref
-reference "fig:RNA-PCA"
-plural "false"
-caps "false"
-noprefix "false"
-
-\end_inset
 
-), no such correction is possible for the noise component: Batch 1 simply
- has substantially more random noise in it, which reduces the statistical
- power for any differential expression tests involving samples in that batch.
- 
 \end_layout
 
-\begin_layout Standard
-\begin_inset Float figure
-wide false
-sideways false
-status collapsed
+\begin_layout Plain Layout
+\begin_inset Flex TODO Note (inline)
+status open
 
 \begin_layout Plain Layout
-\align center
-\begin_inset Graphics
-	filename graphics/CD4-csaw/RNA-seq/PCA-final-12-CROP.png
-	lyxscale 25
-	width 100col%
-	groupId colwidth-raster
+Get the IDR threshold
+\end_layout
 
 \end_inset
 
@@ -5441,7 +5833,7 @@ status collapsed
 status collapsed
 
 \begin_layout Plain Layout
-PCoA plot of RNA-seq samples after ComBat batch correction.
+Summary of peak-calling statistics.
 \end_layout
 
 \end_inset
@@ -5449,28 +5841,18 @@ PCoA plot of RNA-seq samples after ComBat batch correction.
 
 \begin_inset CommandInset label
 LatexCommand label
-name "fig:rna-pca-final"
+name "tab:peak-calling-summary"
 
 \end_inset
 
 
 \series bold
-PCoA plot of RNA-seq samples after ComBat batch correction.
+Summary of peak-calling statistics.
  
 \series default
-Each point represents an individual sample.
- Samples with the same combination of cell type and time point are encircled
- with a shaded region to aid in visual identification of the sample groups.
- Samples with of same cell type from the same donor are connected by lines
- to indicate the 
-\begin_inset Quotes eld
-\end_inset
-
-trajectory
-\begin_inset Quotes erd
-\end_inset
-
- of each donor's cells over time in PCoA space.
+For each histone mark, the number of peaks called using SICER at an IDR
+ threshold of ???, the mean width of those peaks, the fraction of the genome
+ covered by peaks, and the fraction of reads in peaks (FRiP).
 \end_layout
 
 \end_inset
@@ -5484,253 +5866,250 @@ trajectory
 \end_layout
 
 \begin_layout Standard
-Despite the difficulty in detecting specific differentially expressed genes,
- there is still evidence that differential expression is present for these
- time points.
- In Figure 
+Table 
 \begin_inset CommandInset ref
 LatexCommand ref
-reference "fig:rna-pca-final"
+reference "tab:peak-calling-summary"
 plural "false"
 caps "false"
 noprefix "false"
 
 \end_inset
 
-, there is a clear separation between naïve and memory samples at Day 0,
- despite the fact that only 2 genes were significantly differentially expressed
- for this comparison.
- Similarly, the small numbers of genes detected for the Day 0 vs Day 5 compariso
-ns do not reflect the large separation between these time points in Figure
- 
+ gives a summary of the peak calling statistics for each histone mark.
+ Consistent with previous observations, all 3 histone marks occur in broad
+ regions spanning many consecutive nucleosomes, rather than in sharp peaks
+ as would be expected for a transcription factor or other molecule that
+ binds to specific sites.
+ This conclusion is further supported by Figure 
 \begin_inset CommandInset ref
 LatexCommand ref
-reference "fig:rna-pca-final"
+reference "fig:CCF-with-blacklist"
 plural "false"
 caps "false"
 noprefix "false"
 
 \end_inset
 
-.
- In addition, the 
-\begin_inset Flex Glossary Term
-status open
+, in which a clear nucleosome-sized periodicity is visible in the cross-correlat
+ion value for each sample, indicating that each time a given mark is present
+ on one histone, it is also likely to be found on adjacent histones as well.
+ H3K27me3 enrichment in particular is substantially more broad than either
+ H3K4 mark, with a mean peak width of almost 19,000 bp.
+ This is also reflected in the periodicity observed in Figure 
+\begin_inset CommandInset ref
+LatexCommand ref
+reference "fig:CCF-with-blacklist"
+plural "false"
+caps "false"
+noprefix "false"
 
-\begin_layout Plain Layout
-MOFA
+\end_inset
+
+, which remains strong much farther out for H3K27me3 than the other marks,
+ showing H3K27me3 especially tends to be found on long runs of consecutive
+ histones.
 \end_layout
 
+\begin_layout Standard
+All 3 histone marks tend to occur more often near promoter regions, as shown
+ in Figure 
+\begin_inset CommandInset ref
+LatexCommand ref
+reference "fig:near-promoter-peak-enrich"
+plural "false"
+caps "false"
+noprefix "false"
+
 \end_inset
 
- 
+.
+ The majority of each density distribution is flat, representing the background
+ density of peaks genome-wide.
+ Each distribution has a peak near zero, representing an enrichment of peaks
+ close to 
 \begin_inset Flex Glossary Term
 status open
 
 \begin_layout Plain Layout
-LF
+TSS
 \end_layout
 
 \end_inset
 
- plots in Figure 
+ positions relative to the remainder of the genome.
+ Interestingly, the 
+\begin_inset Quotes eld
+\end_inset
+
+radius
+\begin_inset Quotes erd
+\end_inset
+
+ within which this enrichment occurs is not the same for every histone mark
+ (Table 
 \begin_inset CommandInset ref
 LatexCommand ref
-reference "fig:mofa-lf-scatter"
+reference "tab:effective-promoter-radius"
 plural "false"
 caps "false"
 noprefix "false"
 
 \end_inset
 
-.
- This suggests that there is indeed a differential expression signal present
- in the data for these comparisons, but the large variability in the Batch
- 1 samples obfuscates this signal at the individual gene level.
- As a result, it is impossible to make any meaningful statements about the
- 
+).
+ For H3K4me2 and H3K4me3, peaks are most enriched within 1
+\begin_inset space ~
+\end_inset
+
+kbp of 
+\begin_inset Flex Glossary Term
+status open
+
+\begin_layout Plain Layout
+TSS
+\end_layout
+
+\end_inset
+
+ positions, while for H3K27me3, enrichment is broader, extending to 2.5
+\begin_inset space ~
+\end_inset
+
+kbp.
+ These 
 \begin_inset Quotes eld
 \end_inset
 
-size
+effective promoter radii
 \begin_inset Quotes erd
 \end_inset
 
- of the gene signature for any time point, since the number of significant
- genes as well as the estimated number of differentially expressed genes
- depends so strongly on the variations in sample quality in addition to
- the size of the differential expression signal in the data.
- Gene-set enrichment analyses are similarly impractical.
- However, analyses looking at genome-wide patterns of expression are still
- practical.
-\end_layout
-
-\begin_layout Subsection
-H3K4 and H3K27 methylation occur in broad regions and are enriched near
- promoters
+ remain approximately the same across all combinations of experimental condition
+ (cell type, time point, and donor), so they appear to be a property of
+ the histone mark itself.
+ Hence, these radii were used to define the promoter regions for each histone
+ mark in all further analyses.
 \end_layout
 
 \begin_layout Standard
-\begin_inset Float table
+\begin_inset Float figure
 wide false
 sideways false
 status open
 
 \begin_layout Plain Layout
-\align center
 \begin_inset Flex TODO Note (inline)
-status open
-
-\begin_layout Plain Layout
-Also get 
-\emph on
-median
-\emph default
- peak width and maybe other quantiles (25%, 75%)
-\end_layout
-
-\end_inset
-
-
-\end_layout
-
-\begin_layout Plain Layout
-\align center
-\begin_inset Tabular
-<lyxtabular version="3" rows="4" columns="5">
-<features tabularvalignment="middle">
-<column alignment="center" valignment="top">
-<column alignment="center" valignment="top">
-<column alignment="center" valignment="top">
-<column alignment="center" valignment="top">
-<column alignment="center" valignment="top">
-<row>
-<cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
-\begin_inset Text
+status open
 
 \begin_layout Plain Layout
-Histone Mark
+Future direction idea: Need a control: shuffle all peaks and repeat, N times.
 \end_layout
 
 \end_inset
-</cell>
-<cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
-\begin_inset Text
 
-\begin_layout Plain Layout
-# Peaks
-\end_layout
 
-\end_inset
-</cell>
-<cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
-\begin_inset Text
+\end_layout
 
 \begin_layout Plain Layout
-Mean peak width
-\end_layout
+\align center
+\begin_inset Graphics
+	filename graphics/CD4-csaw/Promoter Peak Distance Profile-PAGE1-CROP.pdf
+	lyxscale 50
+	width 80col%
 
 \end_inset
-</cell>
-<cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
-\begin_inset Text
 
-\begin_layout Plain Layout
-genome coverage
-\end_layout
 
-\end_inset
-</cell>
-<cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" rightline="true" usebox="none">
-\begin_inset Text
+\end_layout
 
 \begin_layout Plain Layout
-FRiP
-\end_layout
+\begin_inset Caption Standard
 
-\end_inset
-</cell>
-</row>
-<row>
-<cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
-\begin_inset Text
+\begin_layout Plain Layout
+\begin_inset Argument 1
+status collapsed
 
 \begin_layout Plain Layout
-H3K4me2
+Enrichment of peaks in promoter neighborhoods.
 \end_layout
 
 \end_inset
-</cell>
-<cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
-\begin_inset Text
 
-\begin_layout Plain Layout
-14965
-\end_layout
+
+\begin_inset CommandInset label
+LatexCommand label
+name "fig:near-promoter-peak-enrich"
 
 \end_inset
-</cell>
-<cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
-\begin_inset Text
 
-\begin_layout Plain Layout
-3970
+
+\series bold
+Enrichment of peaks in promoter neighborhoods.
+ 
+\series default
+This plot shows the distribution of distances from each annotated transcription
+ start site in the genome to the nearest called peak.
+ Each line represents one combination of histone mark, cell type, and time
+ point.
+ Distributions are smoothed using kernel density estimation.
+ TSSs that occur 
+\emph on
+within
+\emph default
+ peaks were excluded from this plot to avoid a large spike at zero that
+ would overshadow the rest of the distribution.
 \end_layout
 
 \end_inset
-</cell>
-<cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
-\begin_inset Text
 
-\begin_layout Plain Layout
-1.92%
+
 \end_layout
 
 \end_inset
-</cell>
-<cell alignment="center" valignment="top" topline="true" leftline="true" rightline="true" usebox="none">
-\begin_inset Text
 
-\begin_layout Plain Layout
-14.2%
+
 \end_layout
 
-\end_inset
-</cell>
-</row>
-<row>
-<cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
-\begin_inset Text
+\begin_layout Standard
+\begin_inset Float table
+wide false
+sideways false
+status collapsed
 
 \begin_layout Plain Layout
-H3K4me3
-\end_layout
-
-\end_inset
-</cell>
-<cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
+\align center
+\begin_inset Tabular
+<lyxtabular version="3" rows="4" columns="2">
+<features tabularvalignment="middle">
+<column alignment="center" valignment="top">
+<column alignment="center" valignment="top">
+<row>
+<cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
 \begin_inset Text
 
 \begin_layout Plain Layout
-6163
+Histone mark
 \end_layout
 
 \end_inset
 </cell>
-<cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
+<cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" rightline="true" usebox="none">
 \begin_inset Text
 
 \begin_layout Plain Layout
-2946
+Effective promoter radius
 \end_layout
 
 \end_inset
 </cell>
+</row>
+<row>
 <cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
 \begin_inset Text
 
 \begin_layout Plain Layout
-0.588%
+H3K4me2
 \end_layout
 
 \end_inset
@@ -5739,45 +6118,38 @@ H3K4me3
 \begin_inset Text
 
 \begin_layout Plain Layout
-6.57%
+1 kb
 \end_layout
 
 \end_inset
 </cell>
 </row>
 <row>
-<cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
-\begin_inset Text
-
-\begin_layout Plain Layout
-H3K27me3
-\end_layout
-
-\end_inset
-</cell>
-<cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
+<cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
 \begin_inset Text
 
 \begin_layout Plain Layout
-18139
+H3K4me3
 \end_layout
 
 \end_inset
 </cell>
-<cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
+<cell alignment="center" valignment="top" topline="true" leftline="true" rightline="true" usebox="none">
 \begin_inset Text
 
 \begin_layout Plain Layout
-18967
+1 kb
 \end_layout
 
 \end_inset
 </cell>
+</row>
+<row>
 <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
 \begin_inset Text
 
 \begin_layout Plain Layout
-11.1%
+H3K27me3
 \end_layout
 
 \end_inset
@@ -5786,7 +6158,7 @@ H3K27me3
 \begin_inset Text
 
 \begin_layout Plain Layout
-22.5%
+2.5 kb
 \end_layout
 
 \end_inset
@@ -5797,19 +6169,6 @@ H3K27me3
 \end_inset
 
 
-\end_layout
-
-\begin_layout Plain Layout
-\begin_inset Flex TODO Note (inline)
-status open
-
-\begin_layout Plain Layout
-Get the IDR threshold
-\end_layout
-
-\end_inset
-
-
 \end_layout
 
 \begin_layout Plain Layout
@@ -5820,7 +6179,7 @@ Get the IDR threshold
 status collapsed
 
 \begin_layout Plain Layout
-Summary of peak-calling statistics.
+Effective promoter radius for each histone mark.
 \end_layout
 
 \end_inset
@@ -5828,18 +6187,28 @@ Summary of peak-calling statistics.
 
 \begin_inset CommandInset label
 LatexCommand label
-name "tab:peak-calling-summary"
+name "tab:effective-promoter-radius"
 
 \end_inset
 
 
 \series bold
-Summary of peak-calling statistics.
- 
+Effective promoter radius for each histone mark.
+
 \series default
-For each histone mark, the number of peaks called using SICER at an IDR
- threshold of ???, the mean width of those peaks, the fraction of the genome
- covered by peaks, and the fraction of reads in peaks (FRiP).
+ These values represent the approximate distance from transcription start
+ site positions within which an excess of peaks are found, as shown in Figure
+ 
+\begin_inset CommandInset ref
+LatexCommand ref
+reference "fig:near-promoter-peak-enrich"
+plural "false"
+caps "false"
+noprefix "false"
+
+\end_inset
+
+.
 \end_layout
 
 \end_inset
@@ -5853,119 +6222,112 @@ For each histone mark, the number of peaks called using SICER at an IDR
 \end_layout
 
 \begin_layout Standard
-Table 
-\begin_inset CommandInset ref
-LatexCommand ref
-reference "tab:peak-calling-summary"
-plural "false"
-caps "false"
-noprefix "false"
+\begin_inset Flex TODO Note (inline)
+status open
+
+\begin_layout Plain Layout
+Consider also showing figure for distance to nearest peak center, and reference
+ median peak size once that is known.
+\end_layout
 
 \end_inset
 
- gives a summary of the peak calling statistics for each histone mark.
- Consistent with previous observations, all 3 histone marks occur in broad
- regions spanning many consecutive nucleosomes, rather than in sharp peaks
- as would be expected for a transcription factor or other molecule that
- binds to specific sites.
- This conclusion is further supported by Figure 
-\begin_inset CommandInset ref
-LatexCommand ref
-reference "fig:CCF-with-blacklist"
-plural "false"
-caps "false"
-noprefix "false"
+
+\end_layout
+
+\begin_layout Subsection
+H3K4 and H3K27 promoter methylation has broadly the expected correlation
+ with gene expression
+\end_layout
+
+\begin_layout Standard
+H3K4me2 and H3K4me2 have previously been reported as activating marks whose
+ presence in a gene's promoter is associated with higher gene expression,
+ while H3K27me3 has been reported as inactivating 
+\begin_inset CommandInset citation
+LatexCommand cite
+key "LaMere2016,LaMere2017"
+literal "false"
 
 \end_inset
 
-, in which a clear nucleosome-sized periodicity is visible in the cross-correlat
-ion value for each sample, indicating that each time a given mark is present
- on one histone, it is also likely to be found on adjacent histones as well.
- H3K27me3 enrichment in particular is substantially more broad than either
- H3K4 mark, with a mean peak width of almost 19,000 bp.
- This is also reflected in the periodicity observed in Figure 
+.
+ The data are consistent with this characterization: genes whose promoters
+ (as defined by the radii for each histone mark listed in 
 \begin_inset CommandInset ref
 LatexCommand ref
-reference "fig:CCF-with-blacklist"
+reference "tab:effective-promoter-radius"
 plural "false"
 caps "false"
 noprefix "false"
 
 \end_inset
 
-, which remains strong much farther out for H3K27me3 than the other marks,
- showing H3K27me3 especially tends to be found on long runs of consecutive
- histones.
-\end_layout
-
-\begin_layout Standard
-\begin_inset Float figure
-wide false
-sideways false
-status open
-
-\begin_layout Plain Layout
-\begin_inset Flex TODO Note (inline)
-status open
-
-\begin_layout Plain Layout
-Ensure this figure uses the peak calls from the new analysis.
-\end_layout
-
+) overlap with a H3K4me2 or H3K4me3 peak tend to have higher expression
+ than those that don't, while H3K27me3 is likewise associated with lower
+ gene expression, as shown in 
+\begin_inset CommandInset ref
+LatexCommand ref
+reference "fig:fpkm-by-peak"
+plural "false"
+caps "false"
+noprefix "false"
+
 \end_inset
 
+.
+ This pattern holds across all combinations of cell type and time point
+ (Welch's 
+\emph on
+t
+\emph default
+-test, all 
+\begin_inset Formula $p\mathrm{-values}\ll2.2\times10^{-16}$
+\end_inset
 
-\end_layout
+).
+ The difference in average 
+\begin_inset Formula $\log_{2}$
+\end_inset
 
-\begin_layout Plain Layout
-\begin_inset Flex TODO Note (inline)
+ 
+\begin_inset Flex Glossary Term
 status open
 
 \begin_layout Plain Layout
-Need a control: shuffle all peaks and repeat, N times.
- Do real vs shuffled control both in a top/bottom arrangement.
+FPKM
 \end_layout
 
 \end_inset
 
+ values when a peak overlaps the promoter is about 
+\begin_inset Formula $+5.67$
+\end_inset
 
-\end_layout
-
-\begin_layout Plain Layout
-\begin_inset Flex TODO Note (inline)
-status open
-
-\begin_layout Plain Layout
-Consider counting TSS inside peaks as negative number indicating how far
- 
-\emph on
-inside
-\emph default
- the peak the TSS is (i.e.
- distance to nearest non-peak area).
-\end_layout
-
+ for H3K4me2, 
+\begin_inset Formula $+5.76$
 \end_inset
 
+ for H3K4me2, and 
+\begin_inset Formula $-4.00$
+\end_inset
 
+ for H3K27me3.
 \end_layout
 
+\begin_layout Standard
+\begin_inset Float figure
+wide false
+sideways false
+status collapsed
+
 \begin_layout Plain Layout
 \begin_inset Flex TODO Note (inline)
 status open
 
 \begin_layout Plain Layout
-The H3K4 part of this figure is included in 
-\begin_inset CommandInset citation
-LatexCommand cite
-key "LaMere2016"
-literal "false"
-
-\end_inset
-
- as Fig.
- S2.
- Do I need to do anything about that?
+This figure is generated from the old analysis.
+ Either note that in some way or re-generate it from the new peak calls.
 \end_layout
 
 \end_inset
@@ -5976,9 +6338,9 @@ literal "false"
 \begin_layout Plain Layout
 \align center
 \begin_inset Graphics
-	filename graphics/CD4-csaw/Promoter Peak Distance Profile-PAGE1-CROP.pdf
+	filename graphics/CD4-csaw/FPKM by Peak Violin Plots-CROP.pdf
 	lyxscale 50
-	width 80col%
+	width 100col%
 
 \end_inset
 
@@ -5993,7 +6355,7 @@ literal "false"
 status collapsed
 
 \begin_layout Plain Layout
-Enrichment of peaks in promoter neighborhoods.
+Expression distributions of genes with and without promoter peaks.
 \end_layout
 
 \end_inset
@@ -6001,26 +6363,13 @@ Enrichment of peaks in promoter neighborhoods.
 
 \begin_inset CommandInset label
 LatexCommand label
-name "fig:near-promoter-peak-enrich"
+name "fig:fpkm-by-peak"
 
 \end_inset
 
 
 \series bold
-Enrichment of peaks in promoter neighborhoods.
- 
-\series default
-This plot shows the distribution of distances from each annotated transcription
- start site in the genome to the nearest called peak.
- Each line represents one combination of histone mark, cell type, and time
- point.
- Distributions are smoothed using kernel density estimation.
- TSSs that occur 
-\emph on
-within
-\emph default
- peaks were excluded from this plot to avoid a large spike at zero that
- would overshadow the rest of the distribution.
+Expression distributions of genes with and without promoter peaks.
 \end_layout
 
 \end_inset
@@ -6033,100 +6382,148 @@ within
 
 \end_layout
 
+\begin_layout Subsection
+Gene expression and promoter histone methylation patterns in naïve and memory
+ show convergence at day 14
+\end_layout
+
 \begin_layout Standard
-\begin_inset Float table
-wide false
-sideways false
-status collapsed
+We hypothesized that if naïve cells had differentiated into memory cells
+ by Day 14, then their patterns of expression and histone modification should
+ converge with those of memory cells at Day 14.
+ Figure 
+\begin_inset CommandInset ref
+LatexCommand ref
+reference "fig:PCoA-promoters"
+plural "false"
+caps "false"
+noprefix "false"
 
-\begin_layout Plain Layout
-\align center
-\begin_inset Tabular
-<lyxtabular version="3" rows="4" columns="2">
-<features tabularvalignment="middle">
-<column alignment="center" valignment="top">
-<column alignment="center" valignment="top">
-<row>
-<cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
-\begin_inset Text
+\end_inset
+
+ shows the patterns of variation in all 3 histone marks in the promoter
+ regions of the genome using 
+\begin_inset Flex Glossary Term
+status open
 
 \begin_layout Plain Layout
-Histone mark
+PCoA
 \end_layout
 
 \end_inset
-</cell>
-<cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" rightline="true" usebox="none">
-\begin_inset Text
 
-\begin_layout Plain Layout
-Effective promoter radius
-\end_layout
+.
+ All 3 marks show a noticeable convergence between the naïve and memory
+ samples at day 14, visible as an overlapping of the day 14 groups on each
+ plot.
+ This is consistent with the counts of significantly differentially modified
+ promoters and estimates of the total numbers of differentially modified
+ promoters shown in Table 
+\begin_inset CommandInset ref
+LatexCommand ref
+reference "tab:Number-signif-promoters"
+plural "false"
+caps "false"
+noprefix "false"
 
 \end_inset
-</cell>
-</row>
-<row>
-<cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
-\begin_inset Text
+
+.
+ For all histone marks, evidence of differential modification between naïve
+ and memory samples was detected at every time point except day 14.
+ The day 14 convergence pattern is also present in the 
+\begin_inset Flex Glossary Term
+status open
 
 \begin_layout Plain Layout
-H3K4me2
+RNA-seq
 \end_layout
 
 \end_inset
-</cell>
-<cell alignment="center" valignment="top" topline="true" leftline="true" rightline="true" usebox="none">
-\begin_inset Text
 
-\begin_layout Plain Layout
-1 kb
-\end_layout
+ data (Figure 
+\begin_inset CommandInset ref
+LatexCommand ref
+reference "fig:RNA-PCA-group"
+plural "false"
+caps "false"
+noprefix "false"
 
 \end_inset
-</cell>
-</row>
-<row>
-<cell alignment="center" valignment="top" topline="true" leftline="true" usebox="none">
-\begin_inset Text
+
+), albeit in the 2nd and 3rd principal coordinates, indicating that it is
+ not the most dominant pattern driving gene expression.
+ Taken together, the data show that promoter histone methylation for these
+ 3 histone marks and RNA expression for naïve and memory cells are most
+ similar at day 14, the furthest time point after activation.
+ 
+\begin_inset Flex Glossary Term
+status open
 
 \begin_layout Plain Layout
-H3K4me3
+MOFA
 \end_layout
 
 \end_inset
-</cell>
-<cell alignment="center" valignment="top" topline="true" leftline="true" rightline="true" usebox="none">
-\begin_inset Text
+
+ was also able to capture this day 14 convergence pattern in 
+\begin_inset Flex Glossary Term
+status open
 
 \begin_layout Plain Layout
-1 kb
+LF
 \end_layout
 
 \end_inset
-</cell>
-</row>
-<row>
-<cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
-\begin_inset Text
 
-\begin_layout Plain Layout
-H3K27me3
-\end_layout
+5 (Figure 
+\begin_inset CommandInset ref
+LatexCommand ref
+reference "fig:mofa-lf-scatter"
+plural "false"
+caps "false"
+noprefix "false"
 
 \end_inset
-</cell>
-<cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" rightline="true" usebox="none">
-\begin_inset Text
+
+), which accounts for shared variation across all 3 histone marks and the
+ 
+\begin_inset Flex Glossary Term
+status open
 
 \begin_layout Plain Layout
-2.5 kb
+RNA-seq
 \end_layout
 
 \end_inset
-</cell>
-</row>
-</lyxtabular>
+
+ data, confirming that this convergence is a coordinated pattern across
+ all 4 data sets.
+ While this observation does not prove that the naïve cells have differentiated
+ into memory cells at Day 14, it is consistent with that hypothesis.
+\end_layout
+
+\begin_layout Standard
+\begin_inset Float figure
+placement p
+wide false
+sideways false
+status collapsed
+
+\begin_layout Plain Layout
+\align center
+\begin_inset Float figure
+wide false
+sideways false
+status collapsed
+
+\begin_layout Plain Layout
+\align center
+\begin_inset Graphics
+	filename graphics/CD4-csaw/ChIP-seq/H3K4me2-promoter-PCA-group-CROP.png
+	lyxscale 25
+	width 45col%
+	groupId pcoa-prom-subfig
 
 \end_inset
 
@@ -6137,41 +6534,41 @@ H3K27me3
 \begin_inset Caption Standard
 
 \begin_layout Plain Layout
-\begin_inset Argument 1
-status collapsed
-
-\begin_layout Plain Layout
-Effective promoter radius for each histone mark.
+
+\series bold
+\begin_inset CommandInset label
+LatexCommand label
+name "fig:PCoA-H3K4me2-prom"
+
+\end_inset
+
+PCoA plot of H3K4me2 promoters, after subtracting surrogate variables
 \end_layout
 
 \end_inset
 
 
-\begin_inset CommandInset label
-LatexCommand label
-name "tab:effective-promoter-radius"
+\end_layout
 
 \end_inset
 
 
-\series bold
-Effective promoter radius for each histone mark.
+\begin_inset space \hfill{}
+\end_inset
 
-\series default
- These values represent the approximate distance from transcription start
- site positions within which an excess of peaks are found, as shown in Figure
- 
-\begin_inset CommandInset ref
-LatexCommand ref
-reference "fig:near-promoter-peak-enrich"
-plural "false"
-caps "false"
-noprefix "false"
 
-\end_inset
+\begin_inset Float figure
+wide false
+sideways false
+status collapsed
 
-.
-\end_layout
+\begin_layout Plain Layout
+\align center
+\begin_inset Graphics
+	filename graphics/CD4-csaw/ChIP-seq/H3K4me3-promoter-PCA-group-CROP.png
+	lyxscale 25
+	width 45col%
+	groupId pcoa-prom-subfig
 
 \end_inset
 
@@ -6179,140 +6576,89 @@ noprefix "false"
 \end_layout
 
 \begin_layout Plain Layout
+\begin_inset Caption Standard
 
-\end_layout
+\begin_layout Plain Layout
 
-\end_inset
+\series bold
+\begin_inset CommandInset label
+LatexCommand label
+name "fig:PCoA-H3K4me3-prom"
 
+\end_inset
 
+PCoA plot of H3K4me3 promoters, after subtracting surrogate variables
 \end_layout
 
-\begin_layout Standard
-All 3 histone marks tend to occur more often near promoter regions, as shown
- in Figure 
-\begin_inset CommandInset ref
-LatexCommand ref
-reference "fig:near-promoter-peak-enrich"
-plural "false"
-caps "false"
-noprefix "false"
-
 \end_inset
 
-.
- The majority of each density distribution is flat, representing the background
- density of peaks genome-wide.
- Each distribution has a peak near zero, representing an enrichment of peaks
- close to 
-\begin_inset Flex Glossary Term
-status open
 
-\begin_layout Plain Layout
-TSS
 \end_layout
 
 \end_inset
 
- positions relative to the remainder of the genome.
- Interestingly, the 
-\begin_inset Quotes eld
-\end_inset
 
-radius
-\begin_inset Quotes erd
-\end_inset
+\end_layout
 
- within which this enrichment occurs is not the same for every histone mark
- (Table 
-\begin_inset CommandInset ref
-LatexCommand ref
-reference "tab:effective-promoter-radius"
-plural "false"
-caps "false"
-noprefix "false"
+\begin_layout Plain Layout
+\align center
+\begin_inset Float figure
+wide false
+sideways false
+status collapsed
 
-\end_inset
+\begin_layout Plain Layout
+\align center
+\begin_inset Graphics
+	filename graphics/CD4-csaw/ChIP-seq/H3K27me3-promoter-PCA-group-CROP.png
+	lyxscale 25
+	width 45col%
+	groupId pcoa-prom-subfig
 
-).
- For H3K4me2 and H3K4me3, peaks are most enriched within 1
-\begin_inset space ~
 \end_inset
 
-kbp of 
-\begin_inset Flex Glossary Term
-status open
 
-\begin_layout Plain Layout
-TSS
 \end_layout
 
-\end_inset
+\begin_layout Plain Layout
+\begin_inset Caption Standard
 
- positions, while for H3K27me3, enrichment is broader, extending to 2.5
-\begin_inset space ~
-\end_inset
+\begin_layout Plain Layout
 
-kbp.
- These 
-\begin_inset Quotes eld
-\end_inset
+\series bold
+\begin_inset CommandInset label
+LatexCommand label
+name "fig:PCoA-H3K27me3-prom"
 
-effective promoter radii
-\begin_inset Quotes erd
 \end_inset
 
- remain approximately the same across all combinations of experimental condition
- (cell type, time point, and donor), so they appear to be a property of
- the histone mark itself.
- Hence, these radii were used to define the promoter regions for each histone
- mark in all further analyses.
+PCoA plot of H3K27me3 promoters, after subtracting surrogate variables
 \end_layout
 
-\begin_layout Standard
-\begin_inset Flex TODO Note (inline)
-status open
+\end_inset
+
 
-\begin_layout Plain Layout
-Consider also showing figure for distance to nearest peak center, and reference
- median peak size once that is known.
 \end_layout
 
 \end_inset
 
 
-\end_layout
+\begin_inset space \hfill{}
+\end_inset
 
-\begin_layout Subsection
-H3K4 and H3K27 promoter methylation has broadly the expected correlation
- with gene expression
-\end_layout
 
-\begin_layout Standard
 \begin_inset Float figure
 wide false
 sideways false
 status collapsed
 
-\begin_layout Plain Layout
-\begin_inset Flex TODO Note (inline)
-status open
-
-\begin_layout Plain Layout
-This figure is generated from the old analysis.
- Either note that in some way or re-generate it from the new peak calls.
-\end_layout
-
-\end_inset
-
-
-\end_layout
-
 \begin_layout Plain Layout
 \align center
 \begin_inset Graphics
-	filename graphics/CD4-csaw/FPKM by Peak Violin Plots-CROP.pdf
-	lyxscale 50
-	width 100col%
+	filename graphics/CD4-csaw/RNA-seq/PCA-final-23-CROP.png
+	lyxscale 25
+	width 45col%
+	groupId pcoa-prom-subfig
 
 \end_inset
 
@@ -6323,25 +6669,15 @@ This figure is generated from the old analysis.
 \begin_inset Caption Standard
 
 \begin_layout Plain Layout
-\begin_inset Argument 1
-status collapsed
-
-\begin_layout Plain Layout
-Expression distributions of genes with and without promoter peaks.
-\end_layout
-
-\end_inset
-
 
+\series bold
 \begin_inset CommandInset label
 LatexCommand label
-name "fig:fpkm-by-peak"
+name "fig:RNA-PCA-group"
 
 \end_inset
 
-
-\series bold
-Expression distributions of genes with and without promoter peaks.
+RNA-seq PCoA showing principal coordinates 2 and 3.
 \end_layout
 
 \end_inset
@@ -6354,84 +6690,39 @@ Expression distributions of genes with and without promoter peaks.
 
 \end_layout
 
-\begin_layout Standard
-H3K4me2 and H3K4me2 have previously been reported as activating marks whose
- presence in a gene's promoter is associated with higher gene expression,
- while H3K27me3 has been reported as inactivating 
-\begin_inset CommandInset citation
-LatexCommand cite
-key "LaMere2016,LaMere2017"
-literal "false"
+\begin_layout Plain Layout
+\begin_inset Caption Standard
 
-\end_inset
+\begin_layout Plain Layout
+\begin_inset Argument 1
+status collapsed
 
-.
- The data are consistent with this characterization: genes whose promoters
- (as defined by the radii for each histone mark listed in 
-\begin_inset CommandInset ref
-LatexCommand ref
-reference "tab:effective-promoter-radius"
-plural "false"
-caps "false"
-noprefix "false"
+\begin_layout Plain Layout
+PCoA plots for promoter ChIP-seq and expression RNA-seq data
+\end_layout
 
 \end_inset
 
-) overlap with a H3K4me2 or H3K4me3 peak tend to have higher expression
- than those that don't, while H3K27me3 is likewise associated with lower
- gene expression, as shown in 
-\begin_inset CommandInset ref
-LatexCommand ref
-reference "fig:fpkm-by-peak"
-plural "false"
-caps "false"
-noprefix "false"
-
-\end_inset
 
-.
- This pattern holds across all combinations of cell type and time point
- (Welch's 
-\emph on
-t
-\emph default
--test, all 
-\begin_inset Formula $p\mathrm{-values}\ll2.2\times10^{-16}$
-\end_inset
+\begin_inset CommandInset label
+LatexCommand label
+name "fig:PCoA-promoters"
 
-).
- The difference in average 
-\begin_inset Formula $\log_{2}$
 \end_inset
 
- 
-\begin_inset Flex Glossary Term
-status open
 
-\begin_layout Plain Layout
-FPKM
+\series bold
+PCoA plots for promoter ChIP-seq and expression RNA-seq data
 \end_layout
 
 \end_inset
 
- values when a peak overlaps the promoter is about 
-\begin_inset Formula $+5.67$
-\end_inset
 
- for H3K4me2, 
-\begin_inset Formula $+5.76$
-\end_inset
+\end_layout
 
- for H3K4me2, and 
-\begin_inset Formula $-4.00$
 \end_inset
 
- for H3K27me3.
-\end_layout
-
-\begin_layout Subsection
-Gene expression and promoter histone methylation patterns in naïve and memory
- show convergence at day 14
+
 \end_layout
 
 \begin_layout Standard
@@ -6461,7 +6752,7 @@ begin{landscape}
 \begin_inset Float table
 wide false
 sideways false
-status open
+status collapsed
 
 \begin_layout Plain Layout
 \align center
@@ -6839,312 +7130,78 @@ Day 14
 \end_inset
 </cell>
 <cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
-\begin_inset Text
-
-\begin_layout Plain Layout
-0
-\end_layout
-
-\end_inset
-</cell>
-<cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
-\begin_inset Text
-
-\begin_layout Plain Layout
-0
-\end_layout
-
-\end_inset
-</cell>
-<cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" rightline="true" usebox="none">
-\begin_inset Text
-
-\begin_layout Plain Layout
-0
-\end_layout
-
-\end_inset
-</cell>
-</row>
-</lyxtabular>
-
-\end_inset
-
-
-\end_layout
-
-\begin_layout Plain Layout
-\begin_inset Caption Standard
-
-\begin_layout Plain Layout
-\begin_inset Argument 1
-status collapsed
-
-\begin_layout Plain Layout
-Number of differentially modified promoters between naïve and memory cells
- at each time point after activation.
-\end_layout
-
-\end_inset
-
-
-\begin_inset CommandInset label
-LatexCommand label
-name "tab:Number-signif-promoters"
-
-\end_inset
-
-
-\series bold
-Number of differentially modified promoters between naïve and memory cells
- at each time point after activation.
- 
-\series default
-This table shows both the number of differentially modified promoters detected
- at a 10% FDR threshold (left half), and the total number of differentially
- modified promoters as estimated using the method of 
-\begin_inset CommandInset citation
-LatexCommand cite
-key "Phipson2013"
-literal "false"
-
-\end_inset
-
- (right half).
-\end_layout
-
-\end_inset
-
-
-\end_layout
-
-\end_inset
-
-
-\end_layout
-
-\begin_layout Standard
-\begin_inset ERT
-status open
-
-\begin_layout Plain Layout
-
-
-\backslash
-end{landscape}
-\end_layout
-
-\begin_layout Plain Layout
-
-}
-\end_layout
-
-\end_inset
-
-
-\end_layout
-
-\begin_layout Standard
-We hypothesized that if naïve cells had differentiated into memory cells
- by Day 14, then their patterns of expression and histone modification should
- converge with those of memory cells at Day 14.
- Figure 
-\begin_inset CommandInset ref
-LatexCommand ref
-reference "fig:PCoA-promoters"
-plural "false"
-caps "false"
-noprefix "false"
-
-\end_inset
-
- shows the patterns of variation in all 3 histone marks in the promoter
- regions of the genome using 
-\begin_inset Flex Glossary Term
-status open
-
-\begin_layout Plain Layout
-PCoA
-\end_layout
-
-\end_inset
-
-.
- All 3 marks show a noticeable convergence between the naïve and memory
- samples at day 14, visible as an overlapping of the day 14 groups on each
- plot.
- This is consistent with the counts of significantly differentially modified
- promoters and estimates of the total numbers of differentially modified
- promoters shown in Table 
-\begin_inset CommandInset ref
-LatexCommand ref
-reference "tab:Number-signif-promoters"
-plural "false"
-caps "false"
-noprefix "false"
-
-\end_inset
-
-.
- For all histone marks, evidence of differential modification between naïve
- and memory samples was detected at every time point except day 14.
- The day 14 convergence pattern is also present in the 
-\begin_inset Flex Glossary Term
-status open
-
-\begin_layout Plain Layout
-RNA-seq
-\end_layout
-
-\end_inset
-
- data (Figure 
-\begin_inset CommandInset ref
-LatexCommand ref
-reference "fig:RNA-PCA-group"
-plural "false"
-caps "false"
-noprefix "false"
-
-\end_inset
-
-), albeit in the 2nd and 3rd principal coordinates, indicating that it is
- not the most dominant pattern driving gene expression.
- Taken together, the data show that promoter histone methylation for these
- 3 histone marks and RNA expression for naïve and memory cells are most
- similar at day 14, the furthest time point after activation.
- 
-\begin_inset Flex Glossary Term
-status open
-
-\begin_layout Plain Layout
-MOFA
-\end_layout
-
-\end_inset
-
- was also able to capture this day 14 convergence pattern in 
-\begin_inset Flex Glossary Term
-status open
-
-\begin_layout Plain Layout
-LF
-\end_layout
-
-\end_inset
-
-5 (Figure 
-\begin_inset CommandInset ref
-LatexCommand ref
-reference "fig:mofa-lf-scatter"
-plural "false"
-caps "false"
-noprefix "false"
-
-\end_inset
-
-), which accounts for shared variation across all 3 histone marks and the
- 
-\begin_inset Flex Glossary Term
-status open
-
-\begin_layout Plain Layout
-RNA-seq
-\end_layout
-
-\end_inset
-
- data, confirming that this convergence is a coordinated pattern across
- all 4 data sets.
- While this observation does not prove that the naïve cells have differentiated
- into memory cells at Day 14, it is consistent with that hypothesis.
-\end_layout
-
-\begin_layout Standard
-\begin_inset Float figure
-placement p
-wide false
-sideways false
-status open
-
-\begin_layout Plain Layout
-\align center
-\begin_inset Float figure
-wide false
-sideways false
-status collapsed
-
-\begin_layout Plain Layout
-\align center
-\begin_inset Graphics
-	filename graphics/CD4-csaw/ChIP-seq/H3K4me2-promoter-PCA-group-CROP.png
-	lyxscale 25
-	width 45col%
-	groupId pcoa-prom-subfig
-
-\end_inset
-
-
-\end_layout
-
-\begin_layout Plain Layout
-\begin_inset Caption Standard
+\begin_inset Text
 
 \begin_layout Plain Layout
-
-\series bold
-\begin_inset CommandInset label
-LatexCommand label
-name "fig:PCoA-H3K4me2-prom"
+0
+\end_layout
 
 \end_inset
+</cell>
+<cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" usebox="none">
+\begin_inset Text
 
-PCoA plot of H3K4me2 promoters, after subtracting surrogate variables
+\begin_layout Plain Layout
+0
 \end_layout
 
 \end_inset
+</cell>
+<cell alignment="center" valignment="top" topline="true" bottomline="true" leftline="true" rightline="true" usebox="none">
+\begin_inset Text
 
-
+\begin_layout Plain Layout
+0
 \end_layout
 
 \end_inset
+</cell>
+</row>
+</lyxtabular>
 
-
-\begin_inset space \hfill{}
 \end_inset
 
 
-\begin_inset Float figure
-wide false
-sideways false
+\end_layout
+
+\begin_layout Plain Layout
+\begin_inset Caption Standard
+
+\begin_layout Plain Layout
+\begin_inset Argument 1
 status collapsed
 
 \begin_layout Plain Layout
-\align center
-\begin_inset Graphics
-	filename graphics/CD4-csaw/ChIP-seq/H3K4me3-promoter-PCA-group-CROP.png
-	lyxscale 25
-	width 45col%
-	groupId pcoa-prom-subfig
+Number of differentially modified promoters between naïve and memory cells
+ at each time point after activation.
+\end_layout
 
 \end_inset
 
 
-\end_layout
+\begin_inset CommandInset label
+LatexCommand label
+name "tab:Number-signif-promoters"
 
-\begin_layout Plain Layout
-\begin_inset Caption Standard
+\end_inset
 
-\begin_layout Plain Layout
 
 \series bold
-\begin_inset CommandInset label
-LatexCommand label
-name "fig:PCoA-H3K4me3-prom"
+Number of differentially modified promoters between naïve and memory cells
+ at each time point after activation.
+ 
+\series default
+This table shows both the number of differentially modified promoters detected
+ at a 10% FDR threshold (left half), and the total number of differentially
+ modified promoters as estimated using the method of 
+\begin_inset CommandInset citation
+LatexCommand cite
+key "Phipson2013"
+literal "false"
 
 \end_inset
 
-PCoA plot of H3K4me3 promoters, after subtracting surrogate variables
+ (right half).
 \end_layout
 
 \end_inset
@@ -7157,177 +7214,256 @@ PCoA plot of H3K4me3 promoters, after subtracting surrogate variables
 
 \end_layout
 
-\begin_layout Plain Layout
-\align center
-\begin_inset Float figure
-wide false
-sideways false
-status collapsed
+\begin_layout Standard
+\begin_inset ERT
+status open
 
 \begin_layout Plain Layout
-\align center
-\begin_inset Graphics
-	filename graphics/CD4-csaw/ChIP-seq/H3K27me3-promoter-PCA-group-CROP.png
-	lyxscale 25
-	width 45col%
-	groupId pcoa-prom-subfig
-
-\end_inset
 
 
+\backslash
+end{landscape}
 \end_layout
 
-\begin_layout Plain Layout
-\begin_inset Caption Standard
-
 \begin_layout Plain Layout
 
-\series bold
-\begin_inset CommandInset label
-LatexCommand label
-name "fig:PCoA-H3K27me3-prom"
+}
+\end_layout
 
 \end_inset
 
-PCoA plot of H3K27me3 promoters, after subtracting surrogate variables
+
 \end_layout
 
-\end_inset
+\begin_layout Subsection
+Effect of H3K4me2 and H3K4me3 promoter coverage upstream vs downstream of
+ TSS
+\end_layout
 
+\begin_layout Standard
+\begin_inset Flex TODO Note (inline)
+status open
 
+\begin_layout Plain Layout
+Need a better section title, for this and the next one.
 \end_layout
 
 \end_inset
 
 
-\begin_inset space \hfill{}
+\end_layout
+
+\begin_layout Standard
+\begin_inset Flex TODO Note (inline)
+status open
+
+\begin_layout Plain Layout
+Make sure use of coverage/abundance/whatever is consistent.
+\end_layout
+
 \end_inset
 
 
-\begin_inset Float figure
-wide false
-sideways false
-status collapsed
+\end_layout
+
+\begin_layout Standard
+\begin_inset Flex TODO Note (inline)
+status open
 
 \begin_layout Plain Layout
-\align center
-\begin_inset Graphics
-	filename graphics/CD4-csaw/RNA-seq/PCA-final-23-CROP.png
-	lyxscale 25
-	width 45col%
-	groupId pcoa-prom-subfig
+For the figures in this section and the next, the group labels are arbitrary,
+ so if time allows, it would be good to manually reorder them in a logical
+ way, e.g.
+ most upstream to most downstream.
+ If this is done, make sure to update the text with the correct group labels.
+\end_layout
 
 \end_inset
 
 
 \end_layout
 
-\begin_layout Plain Layout
-\begin_inset Caption Standard
+\begin_layout Standard
+To test whether the position of a histone mark relative to a gene's 
+\begin_inset Flex Glossary Term
+status open
 
 \begin_layout Plain Layout
-
-\series bold
-\begin_inset CommandInset label
-LatexCommand label
-name "fig:RNA-PCA-group"
+TSS
+\end_layout
 
 \end_inset
 
-RNA-seq PCoA showing principal coordinates 2 and 3.
-\end_layout
+ was important, we looked at the 
+\begin_inset Quotes eld
+\end_inset
 
+landscape
+\begin_inset Quotes erd
 \end_inset
 
+ of 
+\begin_inset Flex Glossary Term
+status open
 
+\begin_layout Plain Layout
+ChIP-seq
 \end_layout
 
 \end_inset
 
+ read coverage in naïve Day 0 samples within 5 kb of each gene's 
+\begin_inset Flex Glossary Term
+status open
 
+\begin_layout Plain Layout
+TSS
 \end_layout
 
-\begin_layout Plain Layout
-\begin_inset Caption Standard
+\end_inset
 
-\begin_layout Plain Layout
-\begin_inset Argument 1
-status collapsed
+ by binning reads into 500-bp windows tiled across each promoter 
+\begin_inset Flex Glossary Term
+status open
 
 \begin_layout Plain Layout
-PCoA plots for promoter ChIP-seq and expression RNA-seq data
+logCPM
 \end_layout
 
 \end_inset
 
+ values were calculated for the bins in each promoter and then the average
+ 
+\begin_inset Flex Glossary Term
+status open
 
-\begin_inset CommandInset label
-LatexCommand label
-name "fig:PCoA-promoters"
+\begin_layout Plain Layout
+logCPM
+\end_layout
 
 \end_inset
 
+ for each promoter's bins was normalized to zero, such that the values represent
+ coverage relative to other regions of the same promoter rather than being
+ proportional to absolute read count.
+ The promoters were then clustered based on the normalized bin abundances
+ using 
+\begin_inset Formula $k$
+\end_inset
 
-\series bold
-PCoA plots for promoter ChIP-seq and expression RNA-seq data
-\end_layout
-
+-means clustering with 
+\begin_inset Formula $K=6$
 \end_inset
 
+.
+ Different values of 
+\begin_inset Formula $K$
+\end_inset
 
+ were also tested, but did not substantially change the interpretation of
+ the data.
 \end_layout
 
+\begin_layout Standard
+For H3K4me2, plotting the average bin abundances for each cluster reveals
+ a simple pattern (Figure 
+\begin_inset CommandInset ref
+LatexCommand ref
+reference "fig:H3K4me2-neighborhood-clusters"
+plural "false"
+caps "false"
+noprefix "false"
+
 \end_inset
 
+): Cluster 5 represents a completely flat promoter coverage profile, likely
+ consisting of genes with no H3K4me2 methylation in the promoter.
+ All the other clusters represent a continuum of peak positions relative
+ to the 
+\begin_inset Flex Glossary Term
+status open
 
+\begin_layout Plain Layout
+TSS
 \end_layout
 
-\begin_layout Subsection
-Effect of H3K4me2 and H3K4me3 promoter coverage upstream vs downstream of
- TSS
-\end_layout
+\end_inset
 
-\begin_layout Standard
-\begin_inset Flex TODO Note (inline)
+.
+ In order from must upstream to most downstream, they are Clusters 6, 4,
+ 3, 1, and 2.
+ There do not appear to be any clusters representing coverage patterns other
+ than lone peaks, such as coverage troughs or double peaks.
+ Next, all promoters were plotted in a 
+\begin_inset Flex Glossary Term
 status open
 
 \begin_layout Plain Layout
-Need a better section title, for this and the next one.
+PCA
 \end_layout
 
 \end_inset
 
+ plot based on the same relative bin abundance data, and colored based on
+ cluster membership (Figure 
+\begin_inset CommandInset ref
+LatexCommand ref
+reference "fig:H3K4me2-neighborhood-pca"
+plural "false"
+caps "false"
+noprefix "false"
 
-\end_layout
+\end_inset
 
-\begin_layout Standard
-\begin_inset Flex TODO Note (inline)
+).
+ The 
+\begin_inset Flex Glossary Term
 status open
 
 \begin_layout Plain Layout
-Make sure use of coverage/abundance/whatever is consistent.
+PCA
 \end_layout
 
 \end_inset
 
+ plot shows Cluster 5 (the 
+\begin_inset Quotes eld
+\end_inset
 
-\end_layout
+no peak
+\begin_inset Quotes erd
+\end_inset
 
-\begin_layout Standard
-\begin_inset Flex TODO Note (inline)
-status open
+ cluster) at the center, with the other clusters arranged in a counter-clockwise
+ arc around it in the order noted above, from most upstream peak to most
+ downstream.
+ Notably, the 
+\begin_inset Quotes eld
+\end_inset
 
-\begin_layout Plain Layout
-For the figures in this section and the next, the group labels are arbitrary,
- so if time allows, it would be good to manually reorder them in a logical
- way, e.g.
- most upstream to most downstream.
- If this is done, make sure to update the text with the correct group labels.
-\end_layout
+clusters
+\begin_inset Quotes erd
+\end_inset
+
+ form a single large 
+\begin_inset Quotes eld
+\end_inset
 
+cloud
+\begin_inset Quotes erd
 \end_inset
 
+ with no apparent separation between them, further supporting the conclusion
+ that these clusters represent an arbitrary partitioning of a continuous
+ distribution of promoter coverage landscapes.
+ While the clusters are a useful abstraction that aids in visualization,
+ they are ultimately not an accurate representation of the data.
+ The continuous nature of the distribution also explains why different values
+ of 
+\begin_inset Formula $K$
+\end_inset
 
+ led to similar conclusions.
 \end_layout
 
 \begin_layout Standard
@@ -7437,7 +7573,6 @@ name "fig:H3K4me2-neighborhood-pca"
 \end_inset
 
 PCA of relative coverage depth, colored by K-means cluster membership.
- 
 \end_layout
 
 \end_inset
@@ -7599,189 +7734,6 @@ end{landscape}
 
 \end_layout
 
-\begin_layout Standard
-To test whether the position of a histone mark relative to a gene's 
-\begin_inset Flex Glossary Term
-status open
-
-\begin_layout Plain Layout
-TSS
-\end_layout
-
-\end_inset
-
- was important, we looked at the 
-\begin_inset Quotes eld
-\end_inset
-
-landscape
-\begin_inset Quotes erd
-\end_inset
-
- of 
-\begin_inset Flex Glossary Term
-status open
-
-\begin_layout Plain Layout
-ChIP-seq
-\end_layout
-
-\end_inset
-
- read coverage in naïve Day 0 samples within 5 kb of each gene's 
-\begin_inset Flex Glossary Term
-status open
-
-\begin_layout Plain Layout
-TSS
-\end_layout
-
-\end_inset
-
- by binning reads into 500-bp windows tiled across each promoter 
-\begin_inset Flex Glossary Term
-status open
-
-\begin_layout Plain Layout
-logCPM
-\end_layout
-
-\end_inset
-
- values were calculated for the bins in each promoter and then the average
- 
-\begin_inset Flex Glossary Term
-status open
-
-\begin_layout Plain Layout
-logCPM
-\end_layout
-
-\end_inset
-
- for each promoter's bins was normalized to zero, such that the values represent
- coverage relative to other regions of the same promoter rather than being
- proportional to absolute read count.
- The promoters were then clustered based on the normalized bin abundances
- using 
-\begin_inset Formula $k$
-\end_inset
-
--means clustering with 
-\begin_inset Formula $K=6$
-\end_inset
-
-.
- Different values of 
-\begin_inset Formula $K$
-\end_inset
-
- were also tested, but did not substantially change the interpretation of
- the data.
-\end_layout
-
-\begin_layout Standard
-For H3K4me2, plotting the average bin abundances for each cluster reveals
- a simple pattern (Figure 
-\begin_inset CommandInset ref
-LatexCommand ref
-reference "fig:H3K4me2-neighborhood-clusters"
-plural "false"
-caps "false"
-noprefix "false"
-
-\end_inset
-
-): Cluster 5 represents a completely flat promoter coverage profile, likely
- consisting of genes with no H3K4me2 methylation in the promoter.
- All the other clusters represent a continuum of peak positions relative
- to the 
-\begin_inset Flex Glossary Term
-status open
-
-\begin_layout Plain Layout
-TSS
-\end_layout
-
-\end_inset
-
-.
- In order from must upstream to most downstream, they are Clusters 6, 4,
- 3, 1, and 2.
- There do not appear to be any clusters representing coverage patterns other
- than lone peaks, such as coverage troughs or double peaks.
- Next, all promoters were plotted in a 
-\begin_inset Flex Glossary Term
-status open
-
-\begin_layout Plain Layout
-PCA
-\end_layout
-
-\end_inset
-
- plot based on the same relative bin abundance data, and colored based on
- cluster membership (Figure 
-\begin_inset CommandInset ref
-LatexCommand ref
-reference "fig:H3K4me2-neighborhood-pca"
-plural "false"
-caps "false"
-noprefix "false"
-
-\end_inset
-
-).
- The 
-\begin_inset Flex Glossary Term
-status open
-
-\begin_layout Plain Layout
-PCA
-\end_layout
-
-\end_inset
-
- plot shows Cluster 5 (the 
-\begin_inset Quotes eld
-\end_inset
-
-no peak
-\begin_inset Quotes erd
-\end_inset
-
- cluster) at the center, with the other clusters arranged in a counter-clockwise
- arc around it in the order noted above, from most upstream peak to most
- downstream.
- Notably, the 
-\begin_inset Quotes eld
-\end_inset
-
-clusters
-\begin_inset Quotes erd
-\end_inset
-
- form a single large 
-\begin_inset Quotes eld
-\end_inset
-
-cloud
-\begin_inset Quotes erd
-\end_inset
-
- with no apparent separation between them, further supporting the conclusion
- that these clusters represent an arbitrary partitioning of a continuous
- distribution of promoter coverage landscapes.
- While the clusters are a useful abstraction that aids in visualization,
- they are ultimately not an accurate representation of the data.
- The continuous nature of the distribution also explains why different values
- of 
-\begin_inset Formula $K$
-\end_inset
-
- led to similar conclusions.
-\end_layout
-
 \begin_layout Standard
 \begin_inset Flex TODO Note (inline)
 status open
@@ -7925,23 +7877,49 @@ radius
 status open
 
 \begin_layout Plain Layout
-TSS
+TSS
+\end_layout
+
+\end_inset
+
+ may have a different degree of influence depending on whether it is upstream
+ or downstream of the 
+\begin_inset Flex Glossary Term
+status open
+
+\begin_layout Plain Layout
+TSS
+\end_layout
+
+\end_inset
+
+.
+\end_layout
+
+\begin_layout Standard
+All observations described above for H3K4me2 
+\begin_inset Flex Glossary Term
+status open
+
+\begin_layout Plain Layout
+ChIP-seq
 \end_layout
 
 \end_inset
 
- may have a different degree of influence depending on whether it is upstream
- or downstream of the 
-\begin_inset Flex Glossary Term
-status open
-
-\begin_layout Plain Layout
-TSS
-\end_layout
+ also appear to hold for H3K4me3 as well (Figure 
+\begin_inset CommandInset ref
+LatexCommand ref
+reference "fig:H3K4me3-neighborhood"
+plural "false"
+caps "false"
+noprefix "false"
 
 \end_inset
 
-.
+).
+ This is expected, since there is a high correlation between the positions
+ where both histone marks occur.
 \end_layout
 
 \begin_layout Standard
@@ -8212,21 +8190,49 @@ end{landscape}
 
 \end_layout
 
+\begin_layout Subsection
+Promoter coverage H3K27me3
+\end_layout
+
 \begin_layout Standard
-All observations described above for H3K4me2 
+Unlike both H3K4 marks, whose main patterns of variation appear directly
+ related to the size and position of a single peak within the promoter,
+ the patterns of H3K27me3 methylation in promoters are more complex (Figure
+ 
+\begin_inset CommandInset ref
+LatexCommand ref
+reference "fig:H3K27me3-neighborhood"
+plural "false"
+caps "false"
+noprefix "false"
+
+\end_inset
+
+).
+ Once again looking at the relative coverage in a 500-bp wide bins in a
+ 5kb radius around each 
 \begin_inset Flex Glossary Term
 status open
 
 \begin_layout Plain Layout
-ChIP-seq
+TSS
 \end_layout
 
 \end_inset
 
- also appear to hold for H3K4me3 as well (Figure 
+, promoters were clustered based on the normalized relative coverage values
+ in each bin using 
+\begin_inset Formula $k$
+\end_inset
+
+-means clustering with 
+\begin_inset Formula $K=6$
+\end_inset
+
+ (Figure 
 \begin_inset CommandInset ref
 LatexCommand ref
-reference "fig:H3K4me3-neighborhood"
+reference "fig:H3K27me3-neighborhood-clusters"
 plural "false"
 caps "false"
 noprefix "false"
@@ -8234,12 +8240,106 @@ noprefix "false"
 \end_inset
 
 ).
- This is expected, since there is a high correlation between the positions
- where both histone marks occur.
+ This time, 3 
+\begin_inset Quotes eld
+\end_inset
+
+axes
+\begin_inset Quotes erd
+\end_inset
+
+ of variation can be observed, each represented by 2 clusters with opposing
+ patterns.
+ The first axis is greater upstream coverage (Cluster 1) vs.
+ greater downstream coverage (Cluster 3); the second axis is the coverage
+ at the 
+\begin_inset Flex Glossary Term
+status open
+
+\begin_layout Plain Layout
+TSS
 \end_layout
 
-\begin_layout Subsection
-Promoter coverage H3K27me3
+\end_inset
+
+ itself: peak (Cluster 4) or trough (Cluster 2); lastly, the third axis
+ represents a trough upstream of the 
+\begin_inset Flex Glossary Term
+status open
+
+\begin_layout Plain Layout
+TSS
+\end_layout
+
+\end_inset
+
+ (Cluster 5) vs.
+ downstream of the 
+\begin_inset Flex Glossary Term
+status open
+
+\begin_layout Plain Layout
+TSS
+\end_layout
+
+\end_inset
+
+ (Cluster 6).
+ Referring to these opposing pairs of clusters as axes of variation is justified
+, because they correspond precisely to the first 3 
+\begin_inset Flex Glossary Term (pl)
+status open
+
+\begin_layout Plain Layout
+PC
+\end_layout
+
+\end_inset
+
+ in the 
+\begin_inset Flex Glossary Term
+status open
+
+\begin_layout Plain Layout
+PCA
+\end_layout
+
+\end_inset
+
+ plot of the relative coverage values (Figure 
+\begin_inset CommandInset ref
+LatexCommand ref
+reference "fig:H3K27me3-neighborhood-pca"
+plural "false"
+caps "false"
+noprefix "false"
+
+\end_inset
+
+).
+ The 
+\begin_inset Flex Glossary Term
+status open
+
+\begin_layout Plain Layout
+PCA
+\end_layout
+
+\end_inset
+
+ plot reveals that as in the case of H3K4me2, all the 
+\begin_inset Quotes eld
+\end_inset
+
+clusters
+\begin_inset Quotes erd
+\end_inset
+
+ are really just sections of a single connected cloud rather than discrete
+ clusters.
+ The cloud is approximately ellipsoid-shaped, with each PC being an axis
+ of the ellipse, and each cluster consisting of a pyramidal section of the
+ ellipsoid.
 \end_layout
 
 \begin_layout Standard
@@ -8469,214 +8569,66 @@ kbp downstream, and the logCPM values were normalized within each promoter
 \begin_inset Formula $K=6$
 \end_inset
 
-,
-\series bold
- 
-\series default
-and the average bin values were plotted for each cluster (a).
- The 
-\begin_inset Formula $x$
-\end_inset
-
--axis is the genomic coordinate of each bin relative to the the transcription
- start site, and the 
-\begin_inset Formula $y$
-\end_inset
-
--axis is the mean relative coverage depth of that bin across all promoters
- in the cluster.
- Each line represents the average 
-\begin_inset Quotes eld
-\end_inset
-
-shape
-\begin_inset Quotes erd
-\end_inset
-
- of the promoter coverage for promoters in that cluster.
- PCA was performed on the same data, and the first two PCs were plotted,
- coloring each point by its K-means cluster identity (b).
- For each cluster, the distribution of gene expression values was plotted
- (c).
-\end_layout
-
-\end_inset
-
-
-\end_layout
-
-\end_inset
-
-
-\end_layout
-
-\begin_layout Standard
-\begin_inset ERT
-status open
-
-\begin_layout Plain Layout
-
-
-\backslash
-end{landscape}
-\end_layout
-
-\begin_layout Plain Layout
-
-}
-\end_layout
-
-\end_inset
-
-
-\end_layout
-
-\begin_layout Standard
-Unlike both H3K4 marks, whose main patterns of variation appear directly
- related to the size and position of a single peak within the promoter,
- the patterns of H3K27me3 methylation in promoters are more complex (Figure
- 
-\begin_inset CommandInset ref
-LatexCommand ref
-reference "fig:H3K27me3-neighborhood"
-plural "false"
-caps "false"
-noprefix "false"
-
-\end_inset
-
-).
- Once again looking at the relative coverage in a 500-bp wide bins in a
- 5kb radius around each 
-\begin_inset Flex Glossary Term
-status open
-
-\begin_layout Plain Layout
-TSS
-\end_layout
-
-\end_inset
-
-, promoters were clustered based on the normalized relative coverage values
- in each bin using 
-\begin_inset Formula $k$
-\end_inset
-
--means clustering with 
-\begin_inset Formula $K=6$
-\end_inset
-
- (Figure 
-\begin_inset CommandInset ref
-LatexCommand ref
-reference "fig:H3K27me3-neighborhood-clusters"
-plural "false"
-caps "false"
-noprefix "false"
+,
+\series bold
+ 
+\series default
+and the average bin values were plotted for each cluster (a).
+ The 
+\begin_inset Formula $x$
+\end_inset
 
+-axis is the genomic coordinate of each bin relative to the the transcription
+ start site, and the 
+\begin_inset Formula $y$
 \end_inset
 
-).
- This time, 3 
+-axis is the mean relative coverage depth of that bin across all promoters
+ in the cluster.
+ Each line represents the average 
 \begin_inset Quotes eld
 \end_inset
 
-axes
+shape
 \begin_inset Quotes erd
 \end_inset
 
- of variation can be observed, each represented by 2 clusters with opposing
- patterns.
- The first axis is greater upstream coverage (Cluster 1) vs.
- greater downstream coverage (Cluster 3); the second axis is the coverage
- at the 
-\begin_inset Flex Glossary Term
-status open
-
-\begin_layout Plain Layout
-TSS
+ of the promoter coverage for promoters in that cluster.
+ PCA was performed on the same data, and the first two PCs were plotted,
+ coloring each point by its K-means cluster identity (b).
+ For each cluster, the distribution of gene expression values was plotted
+ (c).
 \end_layout
 
 \end_inset
 
- itself: peak (Cluster 4) or trough (Cluster 2); lastly, the third axis
- represents a trough upstream of the 
-\begin_inset Flex Glossary Term
-status open
 
-\begin_layout Plain Layout
-TSS
 \end_layout
 
 \end_inset
 
- (Cluster 5) vs.
- downstream of the 
-\begin_inset Flex Glossary Term
-status open
 
-\begin_layout Plain Layout
-TSS
 \end_layout
 
-\end_inset
-
- (Cluster 6).
- Referring to these opposing pairs of clusters as axes of variation is justified
-, because they correspond precisely to the first 3 
-\begin_inset Flex Glossary Term (pl)
+\begin_layout Standard
+\begin_inset ERT
 status open
 
 \begin_layout Plain Layout
-PC
-\end_layout
-
-\end_inset
 
- in the 
-\begin_inset Flex Glossary Term
-status open
 
-\begin_layout Plain Layout
-PCA
+\backslash
+end{landscape}
 \end_layout
 
-\end_inset
-
- plot of the relative coverage values (Figure 
-\begin_inset CommandInset ref
-LatexCommand ref
-reference "fig:H3K27me3-neighborhood-pca"
-plural "false"
-caps "false"
-noprefix "false"
-
-\end_inset
-
-).
- The 
-\begin_inset Flex Glossary Term
-status open
-
 \begin_layout Plain Layout
-PCA
-\end_layout
 
-\end_inset
+}
+\end_layout
 
- plot reveals that as in the case of H3K4me2, all the 
-\begin_inset Quotes eld
 \end_inset
 
-clusters
-\begin_inset Quotes erd
-\end_inset
 
- are really just sections of a single connected cloud rather than discrete
- clusters.
- The cloud is approximately ellipsoid-shaped, with each PC being an axis
- of the ellipse, and each cluster consisting of a pyramidal section of the
- ellipsoid.
 \end_layout
 
 \begin_layout Standard
@@ -9077,83 +9029,6 @@ LF
  would not be expected to converge in this way after activation.
 \end_layout
 
-\begin_layout Standard
-\begin_inset Float figure
-wide false
-sideways false
-status open
-
-\begin_layout Plain Layout
-\align center
-\begin_inset Graphics
-	filename graphics/CD4-csaw/LaMere2016_fig8.pdf
-	lyxscale 50
-	width 60col%
-	groupId colwidth
-
-\end_inset
-
-
-\end_layout
-
-\begin_layout Plain Layout
-\begin_inset Caption Standard
-
-\begin_layout Plain Layout
-\begin_inset Argument 1
-status collapsed
-
-\begin_layout Plain Layout
-Lamere 2016 Figure 8 “Model for the role of H3K4 methylation during CD4
- T-cell activation.
-\begin_inset Quotes erd
-\end_inset
-
-
-\end_layout
-
-\end_inset
-
-
-\begin_inset CommandInset label
-LatexCommand label
-name "fig:Lamere2016-Fig8"
-
-\end_inset
-
-
-\series bold
-Lamere 2016 Figure 8 
-\begin_inset CommandInset citation
-LatexCommand cite
-key "LaMere2016"
-literal "false"
-
-\end_inset
-
-, 
-\begin_inset Quotes eld
-\end_inset
-
-Model for the role of H3K4 methylation during CD4 T-cell activation.
-\begin_inset Quotes erd
-\end_inset
-
- 
-\series default
-Reproduced with permission.
-\end_layout
-
-\end_inset
-
-
-\end_layout
-
-\end_inset
-
-
-\end_layout
-
 \begin_layout Standard
 In H3K4me2, H3K4me3, and 
 \begin_inset Flex Glossary Term
@@ -9252,14 +9127,91 @@ SVA
 \begin_inset Flex Glossary Term
 status open
 
-\begin_layout Plain Layout
-PCoA
+\begin_layout Plain Layout
+PCoA
+\end_layout
+
+\end_inset
+
+ to reveal interesting behaviors in the data that were previously only detectabl
+e by a detailed manual analysis.
+\end_layout
+
+\begin_layout Standard
+\begin_inset Float figure
+wide false
+sideways false
+status open
+
+\begin_layout Plain Layout
+\align center
+\begin_inset Graphics
+	filename graphics/CD4-csaw/LaMere2016_fig8.pdf
+	lyxscale 50
+	width 60col%
+	groupId colwidth
+
+\end_inset
+
+
+\end_layout
+
+\begin_layout Plain Layout
+\begin_inset Caption Standard
+
+\begin_layout Plain Layout
+\begin_inset Argument 1
+status collapsed
+
+\begin_layout Plain Layout
+Lamere 2016 Figure 8 “Model for the role of H3K4 methylation during CD4
+ T-cell activation.
+\begin_inset Quotes erd
+\end_inset
+
+
+\end_layout
+
+\end_inset
+
+
+\begin_inset CommandInset label
+LatexCommand label
+name "fig:Lamere2016-Fig8"
+
+\end_inset
+
+
+\series bold
+Lamere 2016 Figure 8 
+\begin_inset CommandInset citation
+LatexCommand cite
+key "LaMere2016"
+literal "false"
+
+\end_inset
+
+, 
+\begin_inset Quotes eld
+\end_inset
+
+Model for the role of H3K4 methylation during CD4 T-cell activation.
+\begin_inset Quotes erd
+\end_inset
+
+ 
+\series default
+Reproduced with permission.
+\end_layout
+
+\end_inset
+
+
 \end_layout
 
 \end_inset
 
- to reveal interesting behaviors in the data that were previously only detectabl
-e by a detailed manual analysis.
+
 \end_layout
 
 \begin_layout Standard
@@ -9464,104 +9416,6 @@ TSS
 Workflow
 \end_layout
 
-\begin_layout Standard
-\begin_inset ERT
-status open
-
-\begin_layout Plain Layout
-
-
-\backslash
-afterpage{
-\end_layout
-
-\begin_layout Plain Layout
-
-
-\backslash
-begin{landscape}
-\end_layout
-
-\end_inset
-
-
-\end_layout
-
-\begin_layout Standard
-\begin_inset Float figure
-wide false
-sideways false
-status open
-
-\begin_layout Plain Layout
-\align center
-\begin_inset Graphics
-	filename graphics/CD4-csaw/rulegraphs/rulegraph-all.pdf
-	lyxscale 50
-	width 100col%
-	height 95theight%
-
-\end_inset
-
-
-\end_layout
-
-\begin_layout Plain Layout
-\begin_inset Caption Standard
-
-\begin_layout Plain Layout
-\begin_inset Argument 1
-status collapsed
-
-\begin_layout Plain Layout
-Dependency graph of steps in reproducible workflow.
-\end_layout
-
-\end_inset
-
-
-\begin_inset CommandInset label
-LatexCommand label
-name "fig:rulegraph"
-
-\end_inset
-
-
-\series bold
-Dependency graph of steps in reproducible workflow.
-\end_layout
-
-\end_inset
-
-
-\end_layout
-
-\end_inset
-
-
-\end_layout
-
-\begin_layout Standard
-\begin_inset ERT
-status open
-
-\begin_layout Plain Layout
-
-
-\backslash
-end{landscape}
-\end_layout
-
-\begin_layout Plain Layout
-
-}
-\end_layout
-
-\end_inset
-
-
-\end_layout
-
 \begin_layout Standard
 The analyses described in this chapter were organized into a reproducible
  workflow using the Snakemake workflow management system 
@@ -9698,6 +9552,104 @@ noprefix "false"
  have completed, thereby automating the entire workflow from start to finish.
 \end_layout
 
+\begin_layout Standard
+\begin_inset ERT
+status open
+
+\begin_layout Plain Layout
+
+
+\backslash
+afterpage{
+\end_layout
+
+\begin_layout Plain Layout
+
+
+\backslash
+begin{landscape}
+\end_layout
+
+\end_inset
+
+
+\end_layout
+
+\begin_layout Standard
+\begin_inset Float figure
+wide false
+sideways false
+status open
+
+\begin_layout Plain Layout
+\align center
+\begin_inset Graphics
+	filename graphics/CD4-csaw/rulegraphs/rulegraph-all.pdf
+	lyxscale 50
+	width 100col%
+	height 95theight%
+
+\end_inset
+
+
+\end_layout
+
+\begin_layout Plain Layout
+\begin_inset Caption Standard
+
+\begin_layout Plain Layout
+\begin_inset Argument 1
+status collapsed
+
+\begin_layout Plain Layout
+Dependency graph of steps in reproducible workflow.
+\end_layout
+
+\end_inset
+
+
+\begin_inset CommandInset label
+LatexCommand label
+name "fig:rulegraph"
+
+\end_inset
+
+
+\series bold
+Dependency graph of steps in reproducible workflow.
+\end_layout
+
+\end_inset
+
+
+\end_layout
+
+\end_inset
+
+
+\end_layout
+
+\begin_layout Standard
+\begin_inset ERT
+status open
+
+\begin_layout Plain Layout
+
+
+\backslash
+end{landscape}
+\end_layout
+
+\begin_layout Plain Layout
+
+}
+\end_layout
+
+\end_inset
+
+
+\end_layout
+
 \begin_layout Standard
 In addition to simply making it easier to organize the steps in the analysis,
  structuring the analysis as a workflow allowed for some analysis strategies
@@ -10678,6 +10630,49 @@ DNA methylation arrays are a relatively new kind of assay that uses microarrays
  by thymidines and interrogates the level of unmethylated DNA.
 \end_layout
 
+\begin_layout Standard
+After normalization, these two probe intensities are summarized in one of
+ two ways, each with advantages and disadvantages.
+ β
+\series bold
+ 
+\series default
+values, interpreted as fraction of DNA copies methylated, range from 0 to
+ 1.
+ β
+\series bold
+ 
+\series default
+values are conceptually easy to interpret, but the constrained range makes
+ them unsuitable for linear modeling, and their error distributions are
+ highly non-normal, which also frustrates linear modeling.
+ M-values, interpreted as the log ratio of methylated to unmethylated copies,
+ are computed by mapping the beta values from 
+\begin_inset Formula $[0,1]$
+\end_inset
+
+ onto 
+\begin_inset Formula $(-\infty,+\infty)$
+\end_inset
+
+ using a sigmoid curve (Figure 
+\begin_inset CommandInset ref
+LatexCommand ref
+reference "fig:Sigmoid-beta-m-mapping"
+plural "false"
+caps "false"
+noprefix "false"
+
+\end_inset
+
+).
+ This transformation results in values with better statistical properties:
+ the unconstrained range is suitable for linear modeling, and the error
+ distributions are more normal.
+ Hence, most linear modeling and other statistical testing on methylation
+ arrays is performed using M-values.
+\end_layout
+
 \begin_layout Standard
 \begin_inset Float figure
 wide false
@@ -10713,18 +10708,13 @@ Sigmoid shape of the mapping between β and M values.
 
 \begin_inset CommandInset label
 LatexCommand label
-name "fig:Sigmoid-beta-m-mapping"
-
-\end_inset
-
-
-\series bold
-Sigmoid shape of the mapping between β and M values.
-\end_layout
+name "fig:Sigmoid-beta-m-mapping"
 
 \end_inset
 
 
+\series bold
+Sigmoid shape of the mapping between β and M values.
 \end_layout
 
 \end_inset
@@ -10732,47 +10722,9 @@ Sigmoid shape of the mapping between β and M values.
 
 \end_layout
 
-\begin_layout Standard
-After normalization, these two probe intensities are summarized in one of
- two ways, each with advantages and disadvantages.
- β
-\series bold
- 
-\series default
-values, interpreted as fraction of DNA copies methylated, range from 0 to
- 1.
- β
-\series bold
- 
-\series default
-values are conceptually easy to interpret, but the constrained range makes
- them unsuitable for linear modeling, and their error distributions are
- highly non-normal, which also frustrates linear modeling.
- M-values, interpreted as the log ratio of methylated to unmethylated copies,
- are computed by mapping the beta values from 
-\begin_inset Formula $[0,1]$
-\end_inset
-
- onto 
-\begin_inset Formula $(-\infty,+\infty)$
 \end_inset
 
- using a sigmoid curve (Figure 
-\begin_inset CommandInset ref
-LatexCommand ref
-reference "fig:Sigmoid-beta-m-mapping"
-plural "false"
-caps "false"
-noprefix "false"
-
-\end_inset
 
-).
- This transformation results in values with better statistical properties:
- the unconstrained range is suitable for linear modeling, and the error
- distributions are more normal.
- Hence, most linear modeling and other statistical testing on methylation
- arrays is performed using M-values.
 \end_layout
 
 \begin_layout Standard
@@ -12028,11 +11980,78 @@ Reconsider subsection organization?
 Separate normalization with RMA introduces unwanted biases in classification
 \end_layout
 
+\begin_layout Standard
+To demonstrate the problem with non-single-channel normalization methods,
+ we considered the problem of training a classifier to distinguish 
+\begin_inset Flex Glossary Term
+status open
+
+\begin_layout Plain Layout
+TX
+\end_layout
+
+\end_inset
+
+ from 
+\begin_inset Flex Glossary Term
+status open
+
+\begin_layout Plain Layout
+AR
+\end_layout
+
+\end_inset
+
+ using the samples from the internal set as training data, evaluating performanc
+e on the external set.
+ First, training and evaluation were performed after normalizing all array
+ samples together as a single set using 
+\begin_inset Flex Glossary Term
+status open
+
+\begin_layout Plain Layout
+RMA
+\end_layout
+
+\end_inset
+
+, and second, the internal samples were normalized separately from the external
+ samples and the training and evaluation were repeated.
+ For each sample in the validation set, the classifier probabilities from
+ both classifiers were plotted against each other (Fig.
+ 
+\begin_inset CommandInset ref
+LatexCommand ref
+reference "fig:Classifier-probabilities-RMA"
+plural "false"
+caps "false"
+noprefix "false"
+
+\end_inset
+
+).
+ As expected, separate normalization biases the classifier probabilities,
+ resulting in several misclassifications.
+ In this case, the bias from separate normalization causes the classifier
+ to assign a lower probability of 
+\begin_inset Flex Glossary Term
+status open
+
+\begin_layout Plain Layout
+AR
+\end_layout
+
+\end_inset
+
+ to every sample.
+ 
+\end_layout
+
 \begin_layout Standard
 \begin_inset Float figure
 wide false
 sideways false
-status open
+status collapsed
 
 \begin_layout Plain Layout
 \align center
@@ -12096,32 +12115,55 @@ The PAM classifier algorithm was trained on the training set of arrays to
 
 \end_layout
 
+\begin_layout Subsection
+fRMA and SCAN maintain classification performance while eliminating dependence
+ on normalization strategy
+\end_layout
+
 \begin_layout Standard
-To demonstrate the problem with non-single-channel normalization methods,
- we considered the problem of training a classifier to distinguish 
+For internal validation, the 6 methods' AUC values ranged from 0.816 to 0.891,
+ as shown in Table 
+\begin_inset CommandInset ref
+LatexCommand ref
+reference "tab:AUC-PAM"
+plural "false"
+caps "false"
+noprefix "false"
+
+\end_inset
+
+.
+ Among the non-single-channel normalizations, dChip outperformed 
 \begin_inset Flex Glossary Term
 status open
 
 \begin_layout Plain Layout
-TX
+RMA
 \end_layout
 
 \end_inset
 
- from 
+, while 
 \begin_inset Flex Glossary Term
 status open
 
 \begin_layout Plain Layout
-AR
+GRSN
 \end_layout
 
 \end_inset
 
- using the samples from the internal set as training data, evaluating performanc
-e on the external set.
- First, training and evaluation were performed after normalizing all array
- samples together as a single set using 
+ reduced the 
+\begin_inset Flex Glossary Term
+status open
+
+\begin_layout Plain Layout
+AUC
+\end_layout
+
+\end_inset
+
+ values for both dChip and 
 \begin_inset Flex Glossary Term
 status open
 
@@ -12131,48 +12173,147 @@ RMA
 
 \end_inset
 
-, and second, the internal samples were normalized separately from the external
- samples and the training and evaluation were repeated.
- For each sample in the validation set, the classifier probabilities from
- both classifiers were plotted against each other (Fig.
- 
+.
+ Both single-channel methods, 
+\begin_inset Flex Glossary Term
+status open
+
+\begin_layout Plain Layout
+fRMA
+\end_layout
+
+\end_inset
+
+ and 
+\begin_inset Flex Glossary Term
+status open
+
+\begin_layout Plain Layout
+SCAN
+\end_layout
+
+\end_inset
+
+, slightly outperformed 
+\begin_inset Flex Glossary Term
+status open
+
+\begin_layout Plain Layout
+RMA
+\end_layout
+
+\end_inset
+
+, with 
+\begin_inset Flex Glossary Term
+status open
+
+\begin_layout Plain Layout
+fRMA
+\end_layout
+
+\end_inset
+
+ ahead of 
+\begin_inset Flex Glossary Term
+status open
+
+\begin_layout Plain Layout
+SCAN
+\end_layout
+
+\end_inset
+
+.
+ However, the difference between 
+\begin_inset Flex Glossary Term
+status open
+
+\begin_layout Plain Layout
+RMA
+\end_layout
+
+\end_inset
+
+ and 
+\begin_inset Flex Glossary Term
+status open
+
+\begin_layout Plain Layout
+fRMA
+\end_layout
+
+\end_inset
+
+ is still quite small.
+ Figure 
 \begin_inset CommandInset ref
 LatexCommand ref
-reference "fig:Classifier-probabilities-RMA"
+reference "fig:ROC-PAM-int"
 plural "false"
 caps "false"
 noprefix "false"
 
 \end_inset
 
-).
- As expected, separate normalization biases the classifier probabilities,
- resulting in several misclassifications.
- In this case, the bias from separate normalization causes the classifier
- to assign a lower probability of 
+ shows that the 
+\begin_inset Flex Glossary Term
+status open
+
+\begin_layout Plain Layout
+ROC
+\end_layout
+
+\end_inset
+
+ curves for 
+\begin_inset Flex Glossary Term
+status open
+
+\begin_layout Plain Layout
+RMA
+\end_layout
+
+\end_inset
+
+, dChip, and 
+\begin_inset Flex Glossary Term
+status open
+
+\begin_layout Plain Layout
+fRMA
+\end_layout
+
+\end_inset
+
+ look very similar and relatively smooth, while both 
 \begin_inset Flex Glossary Term
 status open
 
 \begin_layout Plain Layout
-AR
+GRSN
 \end_layout
 
 \end_inset
 
- to every sample.
- 
+ curves and the curve for 
+\begin_inset Flex Glossary Term
+status open
+
+\begin_layout Plain Layout
+SCAN
 \end_layout
 
-\begin_layout Subsection
-fRMA and SCAN maintain classification performance while eliminating dependence
- on normalization strategy
+\end_inset
+
+ have a more jagged appearance.
 \end_layout
 
 \begin_layout Standard
 \begin_inset Float figure
 wide false
 sideways false
-status open
+status collapsed
 
 \begin_layout Plain Layout
 \align center
@@ -12307,7 +12448,7 @@ ROC curves were generated for PAM classification of AR vs TX after 6 different
 \begin_inset Float table
 wide false
 sideways false
-status open
+status collapsed
 
 \begin_layout Plain Layout
 \align center
@@ -12904,49 +13045,39 @@ noprefix "false"
 \end_layout
 
 \begin_layout Standard
-For internal validation, the 6 methods' AUC values ranged from 0.816 to 0.891,
- as shown in Table 
-\begin_inset CommandInset ref
-LatexCommand ref
-reference "tab:AUC-PAM"
-plural "false"
-caps "false"
-noprefix "false"
-
-\end_inset
-
-.
- Among the non-single-channel normalizations, dChip outperformed 
+For external validation, as expected, all the 
 \begin_inset Flex Glossary Term
 status open
 
 \begin_layout Plain Layout
-RMA
+AUC
 \end_layout
 
 \end_inset
 
-, while 
-\begin_inset Flex Glossary Term
-status open
-
-\begin_layout Plain Layout
-GRSN
-\end_layout
+ values are lower than the internal validations, ranging from 0.642 to 0.750
+ (Table 
+\begin_inset CommandInset ref
+LatexCommand ref
+reference "tab:AUC-PAM"
+plural "false"
+caps "false"
+noprefix "false"
 
 \end_inset
 
- reduced the 
+).
+ With or without 
 \begin_inset Flex Glossary Term
 status open
 
 \begin_layout Plain Layout
-AUC
+GRSN
 \end_layout
 
 \end_inset
 
- values for both dChip and 
+, 
 \begin_inset Flex Glossary Term
 status open
 
@@ -12956,28 +13087,29 @@ RMA
 
 \end_inset
 
-.
- Both single-channel methods, 
+ shows its dominance over dChip in this more challenging test.
+ Unlike in the internal validation, 
 \begin_inset Flex Glossary Term
 status open
 
 \begin_layout Plain Layout
-fRMA
+GRSN
 \end_layout
 
 \end_inset
 
- and 
+ actually improves the classifier performance for 
 \begin_inset Flex Glossary Term
 status open
 
 \begin_layout Plain Layout
-SCAN
+RMA
 \end_layout
 
 \end_inset
 
-, slightly outperformed 
+, although it does not for dChip.
+ Once again, both single-channel methods perform about on par with 
 \begin_inset Flex Glossary Term
 status open
 
@@ -12997,7 +13129,7 @@ fRMA
 
 \end_inset
 
- ahead of 
+ performing slightly better and 
 \begin_inset Flex Glossary Term
 status open
 
@@ -13007,39 +13139,18 @@ SCAN
 
 \end_inset
 
-.
- However, the difference between 
-\begin_inset Flex Glossary Term
-status open
-
-\begin_layout Plain Layout
-RMA
-\end_layout
-
-\end_inset
-
- and 
-\begin_inset Flex Glossary Term
-status open
-
-\begin_layout Plain Layout
-fRMA
-\end_layout
-
-\end_inset
-
- is still quite small.
+ performing a bit worse.
  Figure 
 \begin_inset CommandInset ref
 LatexCommand ref
-reference "fig:ROC-PAM-int"
+reference "fig:ROC-PAM-ext"
 plural "false"
 caps "false"
 noprefix "false"
 
 \end_inset
 
- shows that the 
+ shows the 
 \begin_inset Flex Glossary Term
 status open
 
@@ -13049,115 +13160,20 @@ ROC
 
 \end_inset
 
- curves for 
-\begin_inset Flex Glossary Term
-status open
-
-\begin_layout Plain Layout
-RMA
-\end_layout
-
-\end_inset
-
-, dChip, and 
-\begin_inset Flex Glossary Term
-status open
-
-\begin_layout Plain Layout
-fRMA
-\end_layout
-
-\end_inset
-
- look very similar and relatively smooth, while both 
-\begin_inset Flex Glossary Term
-status open
-
-\begin_layout Plain Layout
-GRSN
-\end_layout
-
-\end_inset
-
- curves and the curve for 
-\begin_inset Flex Glossary Term
-status open
-
-\begin_layout Plain Layout
-SCAN
-\end_layout
-
-\end_inset
-
- have a more jagged appearance.
-\end_layout
-
-\begin_layout Standard
-For external validation, as expected, all the 
-\begin_inset Flex Glossary Term
-status open
-
-\begin_layout Plain Layout
-AUC
-\end_layout
-
-\end_inset
-
- values are lower than the internal validations, ranging from 0.642 to 0.750
- (Table 
-\begin_inset CommandInset ref
-LatexCommand ref
-reference "tab:AUC-PAM"
-plural "false"
-caps "false"
-noprefix "false"
-
-\end_inset
-
-).
- With or without 
-\begin_inset Flex Glossary Term
-status open
-
-\begin_layout Plain Layout
-GRSN
-\end_layout
-
-\end_inset
-
-, 
-\begin_inset Flex Glossary Term
-status open
-
-\begin_layout Plain Layout
-RMA
-\end_layout
-
-\end_inset
-
- shows its dominance over dChip in this more challenging test.
- Unlike in the internal validation, 
-\begin_inset Flex Glossary Term
-status open
-
-\begin_layout Plain Layout
-GRSN
-\end_layout
-
-\end_inset
-
- actually improves the classifier performance for 
+ curves for the external validation test.
+ As expected, none of them are as clean-looking as the internal validation
+ 
 \begin_inset Flex Glossary Term
 status open
 
 \begin_layout Plain Layout
-RMA
+ROC
 \end_layout
 
 \end_inset
 
-, although it does not for dChip.
- Once again, both single-channel methods perform about on par with 
+ curves.
+ The curves for 
 \begin_inset Flex Glossary Term
 status open
 
@@ -13167,7 +13183,7 @@ RMA
 
 \end_inset
 
-, with 
+, RMA+GRSN, and 
 \begin_inset Flex Glossary Term
 status open
 
@@ -13177,83 +13193,94 @@ fRMA
 
 \end_inset
 
- performing slightly better and 
-\begin_inset Flex Glossary Term
-status open
-
-\begin_layout Plain Layout
-SCAN
+ all look similar, while the other curves look more divergent.
 \end_layout
 
-\end_inset
-
- performing a bit worse.
- Figure 
-\begin_inset CommandInset ref
-LatexCommand ref
-reference "fig:ROC-PAM-ext"
-plural "false"
-caps "false"
-noprefix "false"
-
-\end_inset
-
- shows the 
-\begin_inset Flex Glossary Term
-status open
-
-\begin_layout Plain Layout
-ROC
+\begin_layout Subsection
+fRMA with custom-generated vectors enables single-channel normalization
+ on hthgu133pluspm platform
 \end_layout
 
-\end_inset
-
- curves for the external validation test.
- As expected, none of them are as clean-looking as the internal validation
- 
+\begin_layout Standard
+In order to enable use of 
 \begin_inset Flex Glossary Term
 status open
 
 \begin_layout Plain Layout
-ROC
+fRMA
 \end_layout
 
 \end_inset
 
- curves.
- The curves for 
+ to normalize hthgu133pluspm, a custom set of 
 \begin_inset Flex Glossary Term
 status open
 
 \begin_layout Plain Layout
-RMA
+fRMA
 \end_layout
 
 \end_inset
 
-, RMA+GRSN, and 
-\begin_inset Flex Glossary Term
-status open
+ vectors was created.
+ First, an appropriate batch size was chosen by looking at the number of
+ batches and number of samples included as a function of batch size (Figure
+ 
+\begin_inset CommandInset ref
+LatexCommand ref
+reference "fig:frmatools-batch-size"
+plural "false"
+caps "false"
+noprefix "false"
 
-\begin_layout Plain Layout
-fRMA
-\end_layout
+\end_inset
+
+).
+ For a given batch size, all batches with fewer samples that the chosen
+ size must be ignored during training, while larger batches must be randomly
+ downsampled to the chosen size.
+ Hence, the number of samples included for a given batch size equals the
+ batch size times the number of batches with at least that many samples.
+ From Figure 
+\begin_inset CommandInset ref
+LatexCommand ref
+reference "fig:batch-size-samples"
+plural "false"
+caps "false"
+noprefix "false"
 
 \end_inset
 
- all look similar, while the other curves look more divergent.
-\end_layout
+, it is apparent that that a batch size of 8 maximizes the number of samples
+ included in training.
+ Increasing the batch size beyond this causes too many smaller batches to
+ be excluded, reducing the total number of samples for both tissue types.
+ However, a batch size of 8 is not necessarily optimal.
+ The article introducing frmaTools concluded that it was highly advantageous
+ to use a smaller batch size in order to include more batches, even at the
+ expense of including fewer total samples in training 
+\begin_inset CommandInset citation
+LatexCommand cite
+key "McCall2011"
+literal "false"
 
-\begin_layout Subsection
-fRMA with custom-generated vectors enables single-channel normalization
- on hthgu133pluspm platform
+\end_inset
+
+.
+ To strike an appropriate balance between more batches and more samples,
+ a batch size of 5 was chosen.
+ For both blood and biopsy samples, this increased the number of batches
+ included by 10, with only a modest reduction in the number of samples compared
+ to a batch size of 8.
+ With a batch size of 5, 26 batches of biopsy samples and 46 batches of
+ blood samples were available.
 \end_layout
 
 \begin_layout Standard
 \begin_inset Float figure
 wide false
 sideways false
-status open
+status collapsed
 
 \begin_layout Plain Layout
 \align center
@@ -13393,7 +13420,7 @@ For batch sizes ranging from 3 to 15, the number of batches (a) and samples
 \end_layout
 
 \begin_layout Standard
-In order to enable use of 
+Since 
 \begin_inset Flex Glossary Term
 status open
 
@@ -13403,7 +13430,14 @@ fRMA
 
 \end_inset
 
- to normalize hthgu133pluspm, a custom set of 
+ training requires equal-size batches, larger batches are downsampled randomly.
+ This introduces a nondeterministic step in the generation of normalization
+ vectors.
+ To show that this randomness does not substantially change the outcome,
+ the random downsampling and subsequent vector learning was repeated 5 times,
+ with a different random seed each time.
+ 20 samples were selected at random as a test set and normalized with each
+ of the 5 sets of 
 \begin_inset Flex Glossary Term
 status open
 
@@ -13413,58 +13447,67 @@ fRMA
 
 \end_inset
 
- vectors was created.
- First, an appropriate batch size was chosen by looking at the number of
- batches and number of samples included as a function of batch size (Figure
- 
+ normalization vectors as well as ordinary RMA, and the normalized expression
+ values were compared across normalizations.
+ Figure 
 \begin_inset CommandInset ref
 LatexCommand ref
-reference "fig:frmatools-batch-size"
+reference "fig:m-bx-violin"
 plural "false"
 caps "false"
 noprefix "false"
 
 \end_inset
 
-).
- For a given batch size, all batches with fewer samples that the chosen
- size must be ignored during training, while larger batches must be randomly
- downsampled to the chosen size.
- Hence, the number of samples included for a given batch size equals the
- batch size times the number of batches with at least that many samples.
- From Figure 
-\begin_inset CommandInset ref
-LatexCommand ref
-reference "fig:batch-size-samples"
-plural "false"
-caps "false"
-noprefix "false"
+ shows a summary of these comparisons for biopsy samples.
+ Comparing RMA to each of the 5 
+\begin_inset Flex Glossary Term
+status open
+
+\begin_layout Plain Layout
+fRMA
+\end_layout
 
 \end_inset
 
-, it is apparent that that a batch size of 8 maximizes the number of samples
- included in training.
- Increasing the batch size beyond this causes too many smaller batches to
- be excluded, reducing the total number of samples for both tissue types.
- However, a batch size of 8 is not necessarily optimal.
- The article introducing frmaTools concluded that it was highly advantageous
- to use a smaller batch size in order to include more batches, even at the
- expense of including fewer total samples in training 
-\begin_inset CommandInset citation
-LatexCommand cite
-key "McCall2011"
-literal "false"
+ normalizations, the distribution of log ratios is somewhat wide, indicating
+ that the normalizations disagree on the expression values of a fair number
+ of probe sets.
+ In contrast, comparisons of 
+\begin_inset Flex Glossary Term
+status open
+
+\begin_layout Plain Layout
+fRMA
+\end_layout
 
 \end_inset
 
-.
- To strike an appropriate balance between more batches and more samples,
- a batch size of 5 was chosen.
- For both blood and biopsy samples, this increased the number of batches
- included by 10, with only a modest reduction in the number of samples compared
- to a batch size of 8.
- With a batch size of 5, 26 batches of biopsy samples and 46 batches of
- blood samples were available.
+ against 
+\begin_inset Flex Glossary Term
+status open
+
+\begin_layout Plain Layout
+fRMA
+\end_layout
+
+\end_inset
+
+, the vast majority of probe sets have very small log ratios, indicating
+ a very high agreement between the normalized values generated by the two
+ normalizations.
+ This shows that the 
+\begin_inset Flex Glossary Term
+status open
+
+\begin_layout Plain Layout
+fRMA
+\end_layout
+
+\end_inset
+
+ normalization's behavior is not very sensitive to the random downsampling
+ of larger batches during training.
 \end_layout
 
 \begin_layout Standard
@@ -13505,7 +13548,6 @@ name "fig:m-bx-violin"
 
 \series bold
 Violin plot of inter-normalization log ratios for biopsy samples.
- 
 \end_layout
 
 \end_inset
@@ -13551,7 +13593,6 @@ name "fig:m-pax-violin"
 
 \series bold
 Violin plot of inter-normalization log ratios for blood samples.
- 
 \end_layout
 
 \end_inset
@@ -13606,24 +13647,44 @@ Each of 20 randomly selected samples was normalized with RMA and with 5
 \end_layout
 
 \begin_layout Standard
-Since 
+Figure 
+\begin_inset CommandInset ref
+LatexCommand ref
+reference "fig:ma-bx-rma-frma"
+plural "false"
+caps "false"
+noprefix "false"
+
+\end_inset
+
+ shows an MA plot of the RMA-normalized values against the fRMA-normalized
+ values for the same probe sets and arrays, corresponding to the first row
+ of Figure 
+\begin_inset CommandInset ref
+LatexCommand ref
+reference "fig:m-bx-violin"
+plural "false"
+caps "false"
+noprefix "false"
+
+\end_inset
+
+.
+ This MA plot shows that not only is there a wide distribution of M-values,
+ but the trend of M-values is dependent on the average normalized intensity.
+ This is expected, since the overall trend represents the differences in
+ the quantile normalization step.
+ When running 
 \begin_inset Flex Glossary Term
 status open
 
 \begin_layout Plain Layout
-fRMA
+RMA
 \end_layout
 
 \end_inset
 
- training requires equal-size batches, larger batches are downsampled randomly.
- This introduces a nondeterministic step in the generation of normalization
- vectors.
- To show that this randomness does not substantially change the outcome,
- the random downsampling and subsequent vector learning was repeated 5 times,
- with a different random seed each time.
- 20 samples were selected at random as a test set and normalized with each
- of the 5 sets of 
+, only the quantiles for these specific 20 arrays are used, while for 
 \begin_inset Flex Glossary Term
 status open
 
@@ -13633,20 +13694,18 @@ fRMA
 
 \end_inset
 
- normalization vectors as well as ordinary RMA, and the normalized expression
- values were compared across normalizations.
+ the quantile distribution is taking from all arrays used in training.
  Figure 
 \begin_inset CommandInset ref
 LatexCommand ref
-reference "fig:m-bx-violin"
+reference "fig:ma-bx-frma-frma"
 plural "false"
 caps "false"
 noprefix "false"
 
 \end_inset
 
- shows a summary of these comparisons for biopsy samples.
- Comparing RMA to each of the 5 
+ shows a similar MA plot comparing 2 different 
 \begin_inset Flex Glossary Term
 status open
 
@@ -13656,20 +13715,54 @@ fRMA
 
 \end_inset
 
- normalizations, the distribution of log ratios is somewhat wide, indicating
- that the normalizations disagree on the expression values of a fair number
- of probe sets.
- In contrast, comparisons of 
-\begin_inset Flex Glossary Term
-status open
+ normalizations, corresponding to the 6th row of Figure 
+\begin_inset CommandInset ref
+LatexCommand ref
+reference "fig:m-bx-violin"
+plural "false"
+caps "false"
+noprefix "false"
 
-\begin_layout Plain Layout
-fRMA
-\end_layout
+\end_inset
+
+.
+ The MA plot is very tightly centered around zero with no visible trend.
+ Figures 
+\begin_inset CommandInset ref
+LatexCommand ref
+reference "fig:m-pax-violin"
+plural "false"
+caps "false"
+noprefix "false"
+
+\end_inset
+
+, 
+\begin_inset CommandInset ref
+LatexCommand ref
+reference "fig:MA-PAX-rma-frma"
+plural "false"
+caps "false"
+noprefix "false"
+
+\end_inset
+
+, and 
+\begin_inset CommandInset ref
+LatexCommand ref
+reference "fig:ma-bx-frma-frma"
+plural "false"
+caps "false"
+noprefix "false"
 
 \end_inset
 
- against 
+ show exactly the same information for the blood samples, once again comparing
+ the normalized expression values between normalizations for all probe sets
+ across 20 randomly selected test arrays.
+ Once again, there is a wider distribution of log ratios between RMA-normalized
+ values and fRMA-normalized, and a much tighter distribution when comparing
+ different 
 \begin_inset Flex Glossary Term
 status open
 
@@ -13679,10 +13772,7 @@ fRMA
 
 \end_inset
 
-, the vast majority of probe sets have very small log ratios, indicating
- a very high agreement between the normalized values generated by the two
- normalizations.
- This shows that the 
+ normalizations to each other, indicating that the 
 \begin_inset Flex Glossary Term
 status open
 
@@ -13692,15 +13782,15 @@ fRMA
 
 \end_inset
 
- normalization's behavior is not very sensitive to the random downsampling
- of larger batches during training.
+ training process is robust to random batch downsampling for the blood samples
+ as well.
 \end_layout
 
 \begin_layout Standard
 \begin_inset Float figure
 wide false
 sideways false
-status open
+status collapsed
 
 \begin_layout Plain Layout
 \align center
@@ -13777,7 +13867,6 @@ name "fig:ma-bx-frma-frma"
 \end_inset
 
 fRMA vs fRMA for biopsy samples.
- 
 \end_layout
 
 \end_inset
@@ -13929,148 +14018,51 @@ fRMA vs fRMA
 
 \end_layout
 
-\begin_layout Standard
-Figure 
-\begin_inset CommandInset ref
-LatexCommand ref
-reference "fig:ma-bx-rma-frma"
-plural "false"
-caps "false"
-noprefix "false"
-
-\end_inset
-
- shows an MA plot of the RMA-normalized values against the fRMA-normalized
- values for the same probe sets and arrays, corresponding to the first row
- of Figure 
-\begin_inset CommandInset ref
-LatexCommand ref
-reference "fig:m-bx-violin"
-plural "false"
-caps "false"
-noprefix "false"
-
-\end_inset
-
-.
- This MA plot shows that not only is there a wide distribution of M-values,
- but the trend of M-values is dependent on the average normalized intensity.
- This is expected, since the overall trend represents the differences in
- the quantile normalization step.
- When running 
-\begin_inset Flex Glossary Term
-status open
-
-\begin_layout Plain Layout
-RMA
-\end_layout
-
-\end_inset
-
-, only the quantiles for these specific 20 arrays are used, while for 
-\begin_inset Flex Glossary Term
-status open
-
-\begin_layout Plain Layout
-fRMA
-\end_layout
-
-\end_inset
-
- the quantile distribution is taking from all arrays used in training.
- Figure 
-\begin_inset CommandInset ref
-LatexCommand ref
-reference "fig:ma-bx-frma-frma"
-plural "false"
-caps "false"
-noprefix "false"
-
-\end_inset
-
- shows a similar MA plot comparing 2 different 
-\begin_inset Flex Glossary Term
-status open
-
-\begin_layout Plain Layout
-fRMA
+\begin_layout Subsection
+SVA, voom, and array weights improve model fit for methylation array data
 \end_layout
 
-\end_inset
-
- normalizations, corresponding to the 6th row of Figure 
-\begin_inset CommandInset ref
-LatexCommand ref
-reference "fig:m-bx-violin"
-plural "false"
-caps "false"
-noprefix "false"
-
-\end_inset
-
-.
- The MA plot is very tightly centered around zero with no visible trend.
- Figures 
-\begin_inset CommandInset ref
-LatexCommand ref
-reference "fig:m-pax-violin"
-plural "false"
-caps "false"
-noprefix "false"
-
-\end_inset
-
-, 
+\begin_layout Standard
+Figure 
 \begin_inset CommandInset ref
 LatexCommand ref
-reference "fig:MA-PAX-rma-frma"
+reference "fig:meanvar-basic"
 plural "false"
 caps "false"
 noprefix "false"
 
 \end_inset
 
-, and 
+ shows the relationship between the mean M-value and the standard deviation
+ calculated for each probe in the methylation array data set.
+ A few features of the data are apparent.
+ First, the data are very strongly bimodal, with peaks in the density around
+ M-values of +4 and -4.
+ These modes correspond to methylation sites that are nearly 100% methylated
+ and nearly 100% unmethylated, respectively.
+ The strong bimodality indicates that a majority of probes interrogate sites
+ that fall into one of these two categories.
+ The points in between these modes represent sites that are either partially
+ methylated in many samples, or are fully methylated in some samples and
+ fully unmethylated in other samples, or some combination.
+ The next visible feature of the data is the W-shaped variance trend.
+ The upticks in the variance trend on either side are expected, based on
+ the sigmoid transformation exaggerating small differences at extreme M-values
+ (Figure 
 \begin_inset CommandInset ref
 LatexCommand ref
-reference "fig:ma-bx-frma-frma"
+reference "fig:Sigmoid-beta-m-mapping"
 plural "false"
 caps "false"
 noprefix "false"
 
 \end_inset
 
- show exactly the same information for the blood samples, once again comparing
- the normalized expression values between normalizations for all probe sets
- across 20 randomly selected test arrays.
- Once again, there is a wider distribution of log ratios between RMA-normalized
- values and fRMA-normalized, and a much tighter distribution when comparing
- different 
-\begin_inset Flex Glossary Term
-status open
-
-\begin_layout Plain Layout
-fRMA
-\end_layout
-
-\end_inset
-
- normalizations to each other, indicating that the 
-\begin_inset Flex Glossary Term
-status open
-
-\begin_layout Plain Layout
-fRMA
-\end_layout
-
-\end_inset
-
- training process is robust to random batch downsampling for the blood samples
- as well.
-\end_layout
-
-\begin_layout Subsection
-SVA, voom, and array weights improve model fit for methylation array data
+).
+ However, the uptick in the center is interesting: it indicates that sites
+ that are not constitutively methylated or unmethylated have a higher variance.
+ This could be a genuine biological effect, or it could be spurious noise
+ that is only observable at sites with varying methylation.
 \end_layout
 
 \begin_layout Standard
@@ -14100,7 +14092,7 @@ begin{landscape}
 \begin_inset Float figure
 wide false
 sideways false
-status open
+status collapsed
 
 \begin_layout Plain Layout
 \begin_inset Flex TODO Note (inline)
@@ -14319,49 +14311,6 @@ end{landscape}
 
 \end_layout
 
-\begin_layout Standard
-Figure 
-\begin_inset CommandInset ref
-LatexCommand ref
-reference "fig:meanvar-basic"
-plural "false"
-caps "false"
-noprefix "false"
-
-\end_inset
-
- shows the relationship between the mean M-value and the standard deviation
- calculated for each probe in the methylation array data set.
- A few features of the data are apparent.
- First, the data are very strongly bimodal, with peaks in the density around
- M-values of +4 and -4.
- These modes correspond to methylation sites that are nearly 100% methylated
- and nearly 100% unmethylated, respectively.
- The strong bimodality indicates that a majority of probes interrogate sites
- that fall into one of these two categories.
- The points in between these modes represent sites that are either partially
- methylated in many samples, or are fully methylated in some samples and
- fully unmethylated in other samples, or some combination.
- The next visible feature of the data is the W-shaped variance trend.
- The upticks in the variance trend on either side are expected, based on
- the sigmoid transformation exaggerating small differences at extreme M-values
- (Figure 
-\begin_inset CommandInset ref
-LatexCommand ref
-reference "fig:Sigmoid-beta-m-mapping"
-plural "false"
-caps "false"
-noprefix "false"
-
-\end_inset
-
-).
- However, the uptick in the center is interesting: it indicates that sites
- that are not constitutively methylated or unmethylated have a higher variance.
- This could be a genuine biological effect, or it could be spurious noise
- that is only observable at sites with varying methylation.
-\end_layout
-
 \begin_layout Standard
 In Figure 
 \begin_inset CommandInset ref
@@ -14416,42 +14365,109 @@ absorbed
 Figure 
 \begin_inset CommandInset ref
 LatexCommand ref
-reference "fig:meanvar-sva-voomaw"
+reference "fig:meanvar-sva-voomaw"
+plural "false"
+caps "false"
+noprefix "false"
+
+\end_inset
+
+ shows the mean-variance trend after fitting the model with the observation
+ weights assigned by voom based on the mean-variance trend shown in Figure
+ 
+\begin_inset CommandInset ref
+LatexCommand ref
+reference "fig:meanvar-sva-aw"
+plural "false"
+caps "false"
+noprefix "false"
+
+\end_inset
+
+.
+ As expected, the weights exactly counteract the trend in the data, resulting
+ in a nearly flat trend centered vertically at 1 (i.e.
+ 0 on the log scale).
+ This shows that the observations with extreme M-values have been appropriately
+ down-weighted to account for the fact that the noise in those observations
+ has been amplified by the non-linear M-value transformation.
+ In turn, this gives relatively more weight to observations in the middle
+ region, which are more likely to correspond to probes measuring interesting
+ biology (not constitutively methylated or unmethylated).
+\end_layout
+
+\begin_layout Standard
+To determine whether any of the known experimental factors had an impact
+ on data quality, the sample quality weights estimated from the data were
+ tested for association with each of the experimental factors (Table 
+\begin_inset CommandInset ref
+LatexCommand ref
+reference "tab:weight-covariate-tests"
 plural "false"
 caps "false"
 noprefix "false"
 
 \end_inset
 
- shows the mean-variance trend after fitting the model with the observation
- weights assigned by voom based on the mean-variance trend shown in Figure
- 
+).
+ Diabetes diagnosis was found to have a potentially significant association
+ with the sample weights, with a t-test p-value of 
+\begin_inset Formula $1.06\times10^{-3}$
+\end_inset
+
+.
+ Figure 
 \begin_inset CommandInset ref
 LatexCommand ref
-reference "fig:meanvar-sva-aw"
+reference "fig:diabetes-sample-weights"
 plural "false"
 caps "false"
 noprefix "false"
 
+\end_inset
+
+ shows the distribution of sample weights grouped by diabetes diagnosis.
+ The samples from patients with 
+\begin_inset Flex Glossary Term
+status open
+
+\begin_layout Plain Layout
+T2D
+\end_layout
+
+\end_inset
+
+ were assigned significantly lower weights than those from patients with
+ 
+\begin_inset Flex Glossary Term
+status open
+
+\begin_layout Plain Layout
+T1D
+\end_layout
+
 \end_inset
 
 .
- As expected, the weights exactly counteract the trend in the data, resulting
- in a nearly flat trend centered vertically at 1 (i.e.
- 0 on the log scale).
- This shows that the observations with extreme M-values have been appropriately
- down-weighted to account for the fact that the noise in those observations
- has been amplified by the non-linear M-value transformation.
- In turn, this gives relatively more weight to observations in the middle
- region, which are more likely to correspond to probes measuring interesting
- biology (not constitutively methylated or unmethylated).
+ This indicates that the 
+\begin_inset Flex Glossary Term
+status open
+
+\begin_layout Plain Layout
+T2D
+\end_layout
+
+\end_inset
+
+ samples had an overall higher variance on average across all probes.
+ 
 \end_layout
 
 \begin_layout Standard
 \begin_inset Float table
 wide false
 sideways false
-status open
+status collapsed
 
 \begin_layout Plain Layout
 \align center
@@ -14675,7 +14691,7 @@ t
 \begin_inset Float figure
 wide false
 sideways false
-status open
+status collapsed
 
 \begin_layout Plain Layout
 \begin_inset Flex TODO Note (inline)
@@ -14744,10 +14760,6 @@ literal "false"
 \end_inset
 
 
-\end_layout
-
-\begin_layout Plain Layout
-
 \end_layout
 
 \end_inset
@@ -14756,77 +14768,102 @@ literal "false"
 \end_layout
 
 \begin_layout Standard
-To determine whether any of the known experimental factors had an impact
- on data quality, the sample quality weights estimated from the data were
- tested for association with each of the experimental factors (Table 
+Table 
 \begin_inset CommandInset ref
 LatexCommand ref
-reference "tab:weight-covariate-tests"
+reference "tab:methyl-num-signif"
 plural "false"
 caps "false"
 noprefix "false"
 
 \end_inset
 
-).
- Diabetes diagnosis was found to have a potentially significant association
- with the sample weights, with a t-test p-value of 
-\begin_inset Formula $1.06\times10^{-3}$
+ shows the number of significantly differentially methylated probes reported
+ by each analysis for each comparison of interest at an 
+\begin_inset Flex Glossary Term
+status open
+
+\begin_layout Plain Layout
+FDR
+\end_layout
+
 \end_inset
 
-.
- Figure 
+ of 10%.
+ As expected, the more elaborate analyses, B and C, report more significant
+ probes than the more basic analysis A, consistent with the conclusions
+ above that the data contain hidden systematic variations that must be modeled.
+ Table 
 \begin_inset CommandInset ref
 LatexCommand ref
-reference "fig:diabetes-sample-weights"
+reference "tab:methyl-est-nonnull"
 plural "false"
 caps "false"
 noprefix "false"
 
 \end_inset
 
- shows the distribution of sample weights grouped by diabetes diagnosis.
- The samples from patients with 
-\begin_inset Flex Glossary Term
-status open
-
-\begin_layout Plain Layout
-T2D
-\end_layout
+ shows the estimated number differentially methylated probes for each test
+ from each analysis.
+ This was computed by estimating the proportion of null hypotheses that
+ were true using the method of 
+\begin_inset CommandInset citation
+LatexCommand cite
+key "Phipson2013Thesis"
+literal "false"
 
 \end_inset
 
- were assigned significantly lower weights than those from patients with
- 
-\begin_inset Flex Glossary Term
-status open
-
-\begin_layout Plain Layout
-T1D
-\end_layout
+ and subtracting that fraction from the total number of probes, yielding
+ an estimate of the number of null hypotheses that are false based on the
+ distribution of p-values across the entire dataset.
+ Note that this does not identify which null hypotheses should be rejected
+ (i.e.
+ which probes are significant); it only estimates the true number of such
+ probes.
+ Once again, analyses B and C result it much larger estimates for the number
+ of differentially methylated probes.
+ In this case, analysis C, the only analysis that includes voom, estimates
+ the largest number of differentially methylated probes for all 3 contrasts.
+ If the assumptions of all the methods employed hold, then this represents
+ a gain in statistical power over the simpler analysis A.
+ Figure 
+\begin_inset CommandInset ref
+LatexCommand ref
+reference "fig:meth-p-value-histograms"
+plural "false"
+caps "false"
+noprefix "false"
 
 \end_inset
 
-.
- This indicates that the 
-\begin_inset Flex Glossary Term
-status open
-
-\begin_layout Plain Layout
-T2D
-\end_layout
+ shows the p-value distributions for each test, from which the numbers in
+ Table 
+\begin_inset CommandInset ref
+LatexCommand ref
+reference "tab:methyl-est-nonnull"
+plural "false"
+caps "false"
+noprefix "false"
 
 \end_inset
 
- samples had an overall higher variance on average across all probes.
- 
+ were generated.
+ The distributions for analysis A all have a dip in density near zero, which
+ is a strong sign of a poor model fit.
+ The histograms for analyses B and C are more well-behaved, with a uniform
+ component stretching all the way from 0 to 1 representing the probes for
+ which the null hypotheses is true (no differential methylation), and a
+ zero-biased component representing the probes for which the null hypothesis
+ is false (differentially methylated).
+ These histograms do not indicate any major issues with the model fit.
 \end_layout
 
 \begin_layout Standard
 \begin_inset Float table
 wide false
 sideways false
-status open
+status collapsed
 
 \begin_layout Plain Layout
 \align center
@@ -15409,10 +15446,6 @@ AR vs.
 \end_inset
 
 
-\end_layout
-
-\begin_layout Plain Layout
-
 \end_layout
 
 \end_inset
@@ -15790,113 +15823,21 @@ literal "false"
 \end_inset
 
 .
- the blue line is only shown in each plot if the estimate of 
-\begin_inset Formula $\hat{\pi}_{0}$
-\end_inset
-
- for that p-value distribution is different from 1.
-\end_layout
-
-\end_inset
-
-
-\end_layout
-
-\end_inset
-
-
-\end_layout
-
-\begin_layout Standard
-Table 
-\begin_inset CommandInset ref
-LatexCommand ref
-reference "tab:methyl-num-signif"
-plural "false"
-caps "false"
-noprefix "false"
-
-\end_inset
-
- shows the number of significantly differentially methylated probes reported
- by each analysis for each comparison of interest at an 
-\begin_inset Flex Glossary Term
-status open
-
-\begin_layout Plain Layout
-FDR
-\end_layout
-
-\end_inset
-
- of 10%.
- As expected, the more elaborate analyses, B and C, report more significant
- probes than the more basic analysis A, consistent with the conclusions
- above that the data contain hidden systematic variations that must be modeled.
- Table 
-\begin_inset CommandInset ref
-LatexCommand ref
-reference "tab:methyl-est-nonnull"
-plural "false"
-caps "false"
-noprefix "false"
-
-\end_inset
-
- shows the estimated number differentially methylated probes for each test
- from each analysis.
- This was computed by estimating the proportion of null hypotheses that
- were true using the method of 
-\begin_inset CommandInset citation
-LatexCommand cite
-key "Phipson2013Thesis"
-literal "false"
-
-\end_inset
-
- and subtracting that fraction from the total number of probes, yielding
- an estimate of the number of null hypotheses that are false based on the
- distribution of p-values across the entire dataset.
- Note that this does not identify which null hypotheses should be rejected
- (i.e.
- which probes are significant); it only estimates the true number of such
- probes.
- Once again, analyses B and C result it much larger estimates for the number
- of differentially methylated probes.
- In this case, analysis C, the only analysis that includes voom, estimates
- the largest number of differentially methylated probes for all 3 contrasts.
- If the assumptions of all the methods employed hold, then this represents
- a gain in statistical power over the simpler analysis A.
- Figure 
-\begin_inset CommandInset ref
-LatexCommand ref
-reference "fig:meth-p-value-histograms"
-plural "false"
-caps "false"
-noprefix "false"
+ the blue line is only shown in each plot if the estimate of 
+\begin_inset Formula $\hat{\pi}_{0}$
+\end_inset
+
+ for that p-value distribution is different from 1.
+\end_layout
 
 \end_inset
 
- shows the p-value distributions for each test, from which the numbers in
- Table 
-\begin_inset CommandInset ref
-LatexCommand ref
-reference "tab:methyl-est-nonnull"
-plural "false"
-caps "false"
-noprefix "false"
+
+\end_layout
 
 \end_inset
 
- were generated.
- The distributions for analysis A all have a dip in density near zero, which
- is a strong sign of a poor model fit.
- The histograms for analyses B and C are more well-behaved, with a uniform
- component stretching all the way from 0 to 1 representing the probes for
- which the null hypotheses is true (no differential methylation), and a
- zero-biased component representing the probes for which the null hypothesis
- is false (differentially methylated).
- These histograms do not indicate any major issues with the model fit.
+
 \end_layout
 
 \begin_layout Standard
@@ -17824,6 +17765,120 @@ Results
 Globin blocking yields a larger and more consistent fraction of useful reads
 \end_layout
 
+\begin_layout Standard
+The objective of the present study was to validate a new protocol for deep
+ 
+\begin_inset Flex Glossary Term
+status open
+
+\begin_layout Plain Layout
+RNA-seq
+\end_layout
+
+\end_inset
+
+ of whole blood drawn into PaxGene tubes from cynomolgus monkeys undergoing
+ islet transplantation, with particular focus on minimizing the loss of
+ useful sequencing space to uninformative globin reads.
+ The details of the analysis with respect to transplant outcomes and the
+ impact of mesenchymal stem cell treatment will be reported in a separate
+ manuscript (in preparation).
+ To focus on the efficacy of our 
+\begin_inset Flex Glossary Term
+status open
+
+\begin_layout Plain Layout
+GB
+\end_layout
+
+\end_inset
+
+ protocol, 37 blood samples, 16 from pre-transplant and 21 from post-transplant
+ time points, were each prepped once with and once without 
+\begin_inset Flex Glossary Term
+status open
+
+\begin_layout Plain Layout
+GB
+\end_layout
+
+\end_inset
+
+ 
+\begin_inset Flex Glossary Term (pl)
+status open
+
+\begin_layout Plain Layout
+oligo
+\end_layout
+
+\end_inset
+
+, and were then sequenced on an Illumina NextSeq500 instrument.
+ The number of reads aligning to each gene in the cynomolgus genome was
+ counted.
+ Table 
+\begin_inset CommandInset ref
+LatexCommand ref
+reference "tab:Fractions-of-reads"
+plural "false"
+caps "false"
+noprefix "false"
+
+\end_inset
+
+ summarizes the distribution of read fractions among the 
+\begin_inset Flex Glossary Term
+status open
+
+\begin_layout Plain Layout
+GB
+\end_layout
+
+\end_inset
+
+ and non-GB libraries.
+ In the libraries with no 
+\begin_inset Flex Glossary Term
+status open
+
+\begin_layout Plain Layout
+GB
+\end_layout
+
+\end_inset
+
+, globin reads made up an average of 44.6% of total input reads, while reads
+ assigned to all other genes made up an average of 26.3%.
+ The remaining reads either aligned to intergenic regions (that include
+ long non-coding RNAs) or did not align with any annotated transcripts in
+ the current build of the cynomolgus genome.
+ In the 
+\begin_inset Flex Glossary Term
+status open
+
+\begin_layout Plain Layout
+GB
+\end_layout
+
+\end_inset
+
+ libraries, globin reads made up only 3.48% and reads assigned to all other
+ genes increased to 50.4%.
+ Thus, 
+\begin_inset Flex Glossary Term
+status open
+
+\begin_layout Plain Layout
+GB
+\end_layout
+
+\end_inset
+
+ resulted in a 92.2% reduction in globin reads and a 91.6% increase in yield
+ of useful non-globin reads.
+\end_layout
+
 \begin_layout Standard
 \begin_inset ERT
 status open
@@ -17852,7 +17907,7 @@ begin{landscape}
 placement p
 wide false
 sideways false
-status open
+status collapsed
 
 \begin_layout Plain Layout
 \align center
@@ -18479,35 +18534,43 @@ end{landscape}
 \end_layout
 
 \begin_layout Standard
-The objective of the present study was to validate a new protocol for deep
+This reduction is not quite as efficient as the previous analysis showed
+ for human samples by DeepSAGE (<0.4% globin reads after globin reduction)
  
+\begin_inset CommandInset citation
+LatexCommand cite
+key "Mastrokolias2012"
+literal "false"
+
+\end_inset
+
+.
+ Nonetheless, this degree of globin reduction is sufficient to nearly double
+ the yield of useful reads.
+ Thus, 
 \begin_inset Flex Glossary Term
 status open
 
 \begin_layout Plain Layout
-RNA-seq
+GB
 \end_layout
 
 \end_inset
 
- of whole blood drawn into PaxGene tubes from cynomolgus monkeys undergoing
- islet transplantation, with particular focus on minimizing the loss of
- useful sequencing space to uninformative globin reads.
- The details of the analysis with respect to transplant outcomes and the
- impact of mesenchymal stem cell treatment will be reported in a separate
- manuscript (in preparation).
- To focus on the efficacy of our 
+ cuts the required sequencing effort (and costs) to achieve a target coverage
+ depth by almost 50%.
+ Consistent with this near doubling of yield, the average difference in
+ un-normalized 
 \begin_inset Flex Glossary Term
 status open
 
 \begin_layout Plain Layout
-GB
+logCPM
 \end_layout
 
 \end_inset
 
- protocol, 37 blood samples, 16 from pre-transplant and 21 from post-transplant
- time points, were each prepped once with and once without 
+ across all genes between the 
 \begin_inset Flex Glossary Term
 status open
 
@@ -18517,20 +18580,24 @@ GB
 
 \end_inset
 
- 
-\begin_inset Flex Glossary Term (pl)
+ libraries and non-GB libraries is approximately 1 (mean = 1.01, median =
+ 1.08), an overall 2-fold increase.
+ Un-normalized values are used here because the 
+\begin_inset Flex Glossary Term
 status open
 
 \begin_layout Plain Layout
-oligo
+TMM
 \end_layout
 
 \end_inset
 
-, and were then sequenced on an Illumina NextSeq500 instrument.
- The number of reads aligning to each gene in the cynomolgus genome was
- counted.
- Table 
+ normalization correctly identifies this 2-fold difference as biologically
+ irrelevant and removes it.
+\end_layout
+
+\begin_layout Standard
+Another important aspect is that the standard deviations in Table 
 \begin_inset CommandInset ref
 LatexCommand ref
 reference "tab:Fractions-of-reads"
@@ -18540,33 +18607,7 @@ noprefix "false"
 
 \end_inset
 
- summarizes the distribution of read fractions among the 
-\begin_inset Flex Glossary Term
-status open
-
-\begin_layout Plain Layout
-GB
-\end_layout
-
-\end_inset
-
- and non-GB libraries.
- In the libraries with no 
-\begin_inset Flex Glossary Term
-status open
-
-\begin_layout Plain Layout
-GB
-\end_layout
-
-\end_inset
-
-, globin reads made up an average of 44.6% of total input reads, while reads
- assigned to all other genes made up an average of 26.3%.
- The remaining reads either aligned to intergenic regions (that include
- long non-coding RNAs) or did not align with any annotated transcripts in
- the current build of the cynomolgus genome.
- In the 
+ are uniformly smaller in the 
 \begin_inset Flex Glossary Term
 status open
 
@@ -18576,9 +18617,11 @@ GB
 
 \end_inset
 
- libraries, globin reads made up only 3.48% and reads assigned to all other
- genes increased to 50.4%.
- Thus, 
+ samples than the non-GB ones, indicating much greater consistency of yield.
+ This is best seen in the percentage of non-globin reads as a fraction of
+ total reads aligned to annotated genes (genic reads).
+ For the non-GB samples, this measure ranges from 10.9% to 80.9%, while for
+ the 
 \begin_inset Flex Glossary Term
 status open
 
@@ -18588,48 +18631,31 @@ GB
 
 \end_inset
 
- resulted in a 92.2% reduction in globin reads and a 91.6% increase in yield
- of useful non-globin reads.
-\end_layout
-
-\begin_layout Standard
-This reduction is not quite as efficient as the previous analysis showed
- for human samples by DeepSAGE (<0.4% globin reads after globin reduction)
- 
-\begin_inset CommandInset citation
-LatexCommand cite
-key "Mastrokolias2012"
-literal "false"
-
-\end_inset
-
-.
- Nonetheless, this degree of globin reduction is sufficient to nearly double
- the yield of useful reads.
- Thus, 
-\begin_inset Flex Glossary Term
-status open
-
-\begin_layout Plain Layout
-GB
-\end_layout
+ samples it ranges from 81.9% to 99.9% (Figure 
+\begin_inset CommandInset ref
+LatexCommand ref
+reference "fig:Fraction-of-genic-reads"
+plural "false"
+caps "false"
+noprefix "false"
 
 \end_inset
 
- cuts the required sequencing effort (and costs) to achieve a target coverage
- depth by almost 50%.
- Consistent with this near doubling of yield, the average difference in
- un-normalized 
+).
+ This means that for applications where it is critical that each sample
+ achieve a specified minimum coverage in order to provide useful information,
+ it would be necessary to budget up to 10 times the sequencing depth per
+ sample without 
 \begin_inset Flex Glossary Term
 status open
 
 \begin_layout Plain Layout
-logCPM
+GB
 \end_layout
 
 \end_inset
 
- across all genes between the 
+, even though the average yield improvement for 
 \begin_inset Flex Glossary Term
 status open
 
@@ -18639,20 +18665,21 @@ GB
 
 \end_inset
 
- libraries and non-GB libraries is approximately 1 (mean = 1.01, median =
- 1.08), an overall 2-fold increase.
- Un-normalized values are used here because the 
+ is only 2-fold, because every sample has a chance of being 90% globin and
+ 10% useful reads.
+ Hence, the more consistent behavior of 
 \begin_inset Flex Glossary Term
 status open
 
 \begin_layout Plain Layout
-TMM
+GB
 \end_layout
 
 \end_inset
 
- normalization correctly identifies this 2-fold difference as biologically
- irrelevant and removes it.
+ samples makes planning an experiment easier and more efficient because
+ it eliminates the need to over-sequence every sample in order to guard
+ against the worst case of a high-globin fraction.
 \end_layout
 
 \begin_layout Standard
@@ -18723,18 +18750,26 @@ Fraction of genic reads in each sample aligned to non-globin genes, with
 
 \end_layout
 
+\begin_layout Subsection
+Globin blocking lowers the noise floor and allows detection of about 2000
+ more low-expression genes
+\end_layout
+
 \begin_layout Standard
-Another important aspect is that the standard deviations in Table 
-\begin_inset CommandInset ref
-LatexCommand ref
-reference "tab:Fractions-of-reads"
-plural "false"
-caps "false"
-noprefix "false"
+\begin_inset Flex TODO Note (inline)
+status open
+
+\begin_layout Plain Layout
+Remove redundant titles from figures
+\end_layout
 
 \end_inset
 
- are uniformly smaller in the 
+
+\end_layout
+
+\begin_layout Standard
+Since 
 \begin_inset Flex Glossary Term
 status open
 
@@ -18744,24 +18779,26 @@ GB
 
 \end_inset
 
- samples than the non-GB ones, indicating much greater consistency of yield.
- This is best seen in the percentage of non-globin reads as a fraction of
- total reads aligned to annotated genes (genic reads).
- For the non-GB samples, this measure ranges from 10.9% to 80.9%, while for
- the 
+ yields more usable sequencing depth, it should also allow detection of
+ more genes at any given threshold.
+ When we looked at the distribution of average normalized 
 \begin_inset Flex Glossary Term
 status open
 
 \begin_layout Plain Layout
-GB
+logCPM
 \end_layout
 
 \end_inset
 
- samples it ranges from 81.9% to 99.9% (Figure 
+ values across all libraries for genes with at least one read assigned to
+ them, we observed the expected bimodal distribution, with a high-abundance
+ "signal" peak representing detected genes and a low-abundance "noise" peak
+ representing genes whose read count did not rise above the noise floor
+ (Figure 
 \begin_inset CommandInset ref
 LatexCommand ref
-reference "fig:Fraction-of-genic-reads"
+reference "fig:logcpm-dists"
 plural "false"
 caps "false"
 noprefix "false"
@@ -18769,10 +18806,8 @@ noprefix "false"
 \end_inset
 
 ).
- This means that for applications where it is critical that each sample
- achieve a specified minimum coverage in order to provide useful information,
- it would be necessary to budget up to 10 times the sequencing depth per
- sample without 
+ Consistent with the 2-fold increase in raw counts assigned to non-globin
+ genes, the signal peak for 
 \begin_inset Flex Glossary Term
 status open
 
@@ -18782,7 +18817,10 @@ GB
 
 \end_inset
 
-, even though the average yield improvement for 
+ samples is shifted to the right relative to the non-GB signal peak.
+ When all the samples are normalized together, this difference is normalized
+ out, lining up the signal peaks, and this reveals that, as expected, the
+ noise floor for the 
 \begin_inset Flex Glossary Term
 status open
 
@@ -18792,9 +18830,8 @@ GB
 
 \end_inset
 
- is only 2-fold, because every sample has a chance of being 90% globin and
- 10% useful reads.
- Hence, the more consistent behavior of 
+ samples is about 2-fold lower.
+ This greater separation between signal and noise peaks in the 
 \begin_inset Flex Glossary Term
 status open
 
@@ -18804,27 +18841,8 @@ GB
 
 \end_inset
 
- samples makes planning an experiment easier and more efficient because
- it eliminates the need to over-sequence every sample in order to guard
- against the worst case of a high-globin fraction.
-\end_layout
-
-\begin_layout Subsection
-Globin blocking lowers the noise floor and allows detection of about 2000
- more low-expression genes
-\end_layout
-
-\begin_layout Standard
-\begin_inset Flex TODO Note (inline)
-status open
-
-\begin_layout Plain Layout
-Remove redundant titles from figures
-\end_layout
-
-\end_inset
-
-
+ samples means that low-expression genes should be more easily detected
+ and more precisely quantified than in the non-GB samples.
 \end_layout
 
 \begin_layout Standard
@@ -18895,10 +18913,6 @@ Distributions of average group gene abundances when normalized separately
 \end_inset
 
 
-\end_layout
-
-\begin_layout Plain Layout
-
 \end_layout
 
 \end_inset
@@ -18907,7 +18921,17 @@ Distributions of average group gene abundances when normalized separately
 \end_layout
 
 \begin_layout Standard
-Since 
+Based on these distributions, we selected a detection threshold of 
+\begin_inset Formula $-1$
+\end_inset
+
+, which is approximately the leftmost edge of the trough between the signal
+ and noise peaks.
+ This represents the most liberal possible detection threshold that doesn't
+ call substantial numbers of noise genes as detected.
+ Among the full dataset, 13429 genes were detected at this threshold, and
+ 22276 were not.
+ When considering the 
 \begin_inset Flex Glossary Term
 status open
 
@@ -18917,35 +18941,20 @@ GB
 
 \end_inset
 
- yields more usable sequencing depth, it should also allow detection of
- more genes at any given threshold.
- When we looked at the distribution of average normalized 
+ libraries and non-GB libraries separately and re-computing normalization
+ factors independently within each group, 14535 genes were detected in the
+ 
 \begin_inset Flex Glossary Term
 status open
 
 \begin_layout Plain Layout
-logCPM
+GB
 \end_layout
 
 \end_inset
 
- values across all libraries for genes with at least one read assigned to
- them, we observed the expected bimodal distribution, with a high-abundance
- "signal" peak representing detected genes and a low-abundance "noise" peak
- representing genes whose read count did not rise above the noise floor
- (Figure 
-\begin_inset CommandInset ref
-LatexCommand ref
-reference "fig:logcpm-dists"
-plural "false"
-caps "false"
-noprefix "false"
-
-\end_inset
-
-).
- Consistent with the 2-fold increase in raw counts assigned to non-globin
- genes, the signal peak for 
+ libraries while only 12460 were detected in the non-GB libraries.
+ Thus, 
 \begin_inset Flex Glossary Term
 status open
 
@@ -18955,10 +18964,8 @@ GB
 
 \end_inset
 
- samples is shifted to the right relative to the non-GB signal peak.
- When all the samples are normalized together, this difference is normalized
- out, lining up the signal peaks, and this reveals that, as expected, the
- noise floor for the 
+ allowed the detection of 2000 extra genes that were buried under the noise
+ floor without 
 \begin_inset Flex Glossary Term
 status open
 
@@ -18968,8 +18975,8 @@ GB
 
 \end_inset
 
- samples is about 2-fold lower.
- This greater separation between signal and noise peaks in the 
+.
+ This pattern of at least 2000 additional genes detected with 
 \begin_inset Flex Glossary Term
 status open
 
@@ -18979,8 +18986,18 @@ GB
 
 \end_inset
 
- samples means that low-expression genes should be more easily detected
- and more precisely quantified than in the non-GB samples.
+ was also consistent across a wide range of possible detection thresholds,
+ from -2 to 3 (see Figure 
+\begin_inset CommandInset ref
+LatexCommand ref
+reference "fig:Gene-detections"
+plural "false"
+caps "false"
+noprefix "false"
+
+\end_inset
+
+).
 \end_layout
 
 \begin_layout Standard
@@ -19052,8 +19069,51 @@ noprefix "false"
 
 \end_layout
 
+\end_inset
+
+
+\end_layout
+
+\begin_layout Subsection
+Globin blocking does not add significant additional noise or decrease sample
+ quality
+\end_layout
+
+\begin_layout Standard
+One potential worry is that the 
+\begin_inset Flex Glossary Term
+status open
+
+\begin_layout Plain Layout
+GB
+\end_layout
+
+\end_inset
+
+ protocol could perturb the levels of non-globin genes.
+ There are two kinds of possible perturbations: systematic and random.
+ The former is not a major concern for detection of differential expression,
+ since a 2-fold change in every sample has no effect on the relative fold
+ change between samples.
+ In contrast, random perturbations would increase the noise and obscure
+ the signal in the dataset, reducing the capacity to detect differential
+ expression.
+\end_layout
+
+\begin_layout Standard
+\begin_inset Flex TODO Note (inline)
+status open
+
 \begin_layout Plain Layout
+Standardize on 
+\begin_inset Quotes eld
+\end_inset
 
+log2
+\begin_inset Quotes erd
+\end_inset
+
+ notation
 \end_layout
 
 \end_inset
@@ -19062,40 +19122,53 @@ noprefix "false"
 \end_layout
 
 \begin_layout Standard
-Based on these distributions, we selected a detection threshold of 
-\begin_inset Formula $-1$
+The data do indeed show small systematic perturbations in gene levels (Figure
+ 
+\begin_inset CommandInset ref
+LatexCommand ref
+reference "fig:MA-plot"
+plural "false"
+caps "false"
+noprefix "false"
+
 \end_inset
 
-, which is approximately the leftmost edge of the trough between the signal
- and noise peaks.
- This represents the most liberal possible detection threshold that doesn't
- call substantial numbers of noise genes as detected.
- Among the full dataset, 13429 genes were detected at this threshold, and
- 22276 were not.
- When considering the 
-\begin_inset Flex Glossary Term
+).
+ Other than the 3 designated alpha and beta globin genes, two other genes
+ stand out as having especially large negative 
+\begin_inset Flex Glossary Term (pl)
 status open
 
 \begin_layout Plain Layout
-GB
+logFC
+\end_layout
+
+\end_inset
+
+: HBD and LOC1021365.
+ HBD, delta globin, is most likely targeted by the blocking 
+\begin_inset Flex Glossary Term (pl)
+status open
+
+\begin_layout Plain Layout
+oligo
 \end_layout
 
 \end_inset
 
- libraries and non-GB libraries separately and re-computing normalization
- factors independently within each group, 14535 genes were detected in the
- 
+ due to high sequence homology with the other globin genes.
+ LOC1021365 is the aforementioned 
 \begin_inset Flex Glossary Term
 status open
 
 \begin_layout Plain Layout
-GB
+ncRNA
 \end_layout
 
 \end_inset
 
- libraries while only 12460 were detected in the non-GB libraries.
- Thus, 
+ that is reverse-complementary to one of the alpha-like genes and that would
+ be expected to be removed during the 
 \begin_inset Flex Glossary Term
 status open
 
@@ -19105,19 +19178,21 @@ GB
 
 \end_inset
 
- allowed the detection of 2000 extra genes that were buried under the noise
- floor without 
+ step.
+ All other genes appear in a cluster centered vertically at 0, and the vast
+ majority of genes in this cluster show an absolute 
 \begin_inset Flex Glossary Term
 status open
 
 \begin_layout Plain Layout
-GB
+logFC
 \end_layout
 
 \end_inset
 
-.
- This pattern of at least 2000 additional genes detected with 
+ of 0.5 or less.
+ Nevertheless, many of these small perturbations are still statistically
+ significant, indicating that the 
 \begin_inset Flex Glossary Term
 status open
 
@@ -19127,44 +19202,18 @@ GB
 
 \end_inset
 
- was also consistent across a wide range of possible detection thresholds,
- from -2 to 3 (see Figure 
-\begin_inset CommandInset ref
-LatexCommand ref
-reference "fig:Gene-detections"
-plural "false"
-caps "false"
-noprefix "false"
-
-\end_inset
-
-).
-\end_layout
-
-\begin_layout Subsection
-Globin blocking does not add significant additional noise or decrease sample
- quality
-\end_layout
-
-\begin_layout Standard
-One potential worry is that the 
-\begin_inset Flex Glossary Term
+ 
+\begin_inset Flex Glossary Term (pl)
 status open
 
 \begin_layout Plain Layout
-GB
+oligo
 \end_layout
 
 \end_inset
 
- protocol could perturb the levels of non-globin genes.
- There are two kinds of possible perturbations: systematic and random.
- The former is not a major concern for detection of differential expression,
- since a 2-fold change in every sample has no effect on the relative fold
- change between samples.
- In contrast, random perturbations would increase the noise and obscure
- the signal in the dataset, reducing the capacity to detect differential
- expression.
+ likely cause very small but non-zero systematic perturbations in measured
+ gene expression levels.
 \end_layout
 
 \begin_layout Standard
@@ -19282,28 +19331,50 @@ edgeR
 status open
 
 \begin_layout Plain Layout
-Standardize on 
-\begin_inset Quotes eld
+Give these numbers the LaTeX math treatment
+\end_layout
+
 \end_inset
 
-log2
-\begin_inset Quotes erd
+
+\end_layout
+
+\begin_layout Standard
+To evaluate the possibility of 
+\begin_inset Flex Glossary Term
+status open
+
+\begin_layout Plain Layout
+GB
+\end_layout
+
 \end_inset
 
- notation
+ causing random perturbations and reducing sample quality, we computed the
+ Pearson correlation between 
+\begin_inset Flex Glossary Term
+status open
+
+\begin_layout Plain Layout
+logCPM
 \end_layout
 
 \end_inset
 
+ values for every pair of samples with and without 
+\begin_inset Flex Glossary Term
+status open
 
+\begin_layout Plain Layout
+GB
 \end_layout
 
-\begin_layout Standard
-The data do indeed show small systematic perturbations in gene levels (Figure
- 
+\end_inset
+
+ and plotted them against each other (Figure 
 \begin_inset CommandInset ref
 LatexCommand ref
-reference "fig:MA-plot"
+reference "fig:gene-abundance-correlations"
 plural "false"
 caps "false"
 noprefix "false"
@@ -19311,41 +19382,54 @@ noprefix "false"
 \end_inset
 
 ).
- Other than the 3 designated alpha and beta globin genes, two other genes
- stand out as having especially large negative 
-\begin_inset Flex Glossary Term (pl)
+ The plot indicated that the 
+\begin_inset Flex Glossary Term
 status open
 
 \begin_layout Plain Layout
-logFC
+GB
 \end_layout
 
 \end_inset
 
-: HBD and LOC1021365.
- HBD, delta globin, is most likely targeted by the blocking 
-\begin_inset Flex Glossary Term (pl)
+ libraries have higher sample-to-sample correlations than the non-GB libraries.
+ Parametric and nonparametric tests for differences between the correlations
+ with and without 
+\begin_inset Flex Glossary Term
 status open
 
 \begin_layout Plain Layout
-oligo
+GB
 \end_layout
 
 \end_inset
 
- due to high sequence homology with the other globin genes.
- LOC1021365 is the aforementioned 
+ both confirmed that this difference was highly significant (2-sided paired
+ t-test: t = 37.2, df = 665, P ≪ 2.2e-16; 2-sided Wilcoxon sign-rank test:
+ V = 2195, P ≪ 2.2e-16).
+ Performing the same tests on the Spearman correlations gave the same conclusion
+ (t-test: t = 26.8, df = 665, P ≪ 2.2e-16; sign-rank test: V = 8781, P ≪ 2.2e-16).
+ The 
+\begin_inset Flex Code
+status open
+
+\begin_layout Plain Layout
+edgeR
+\end_layout
+
+\end_inset
+
+ package was used to compute the overall 
 \begin_inset Flex Glossary Term
 status open
 
 \begin_layout Plain Layout
-ncRNA
+BCV
 \end_layout
 
 \end_inset
 
- that is reverse-complementary to one of the alpha-like genes and that would
- be expected to be removed during the 
+ for 
 \begin_inset Flex Glossary Term
 status open
 
@@ -19355,42 +19439,72 @@ GB
 
 \end_inset
 
- step.
- All other genes appear in a cluster centered vertically at 0, and the vast
- majority of genes in this cluster show an absolute 
+ and non-GB libraries, and found that 
 \begin_inset Flex Glossary Term
 status open
 
 \begin_layout Plain Layout
-logFC
+GB
 \end_layout
 
 \end_inset
 
- of 0.5 or less.
- Nevertheless, many of these small perturbations are still statistically
- significant, indicating that the 
+ resulted in a negligible increase in the 
 \begin_inset Flex Glossary Term
 status open
 
 \begin_layout Plain Layout
-GB
+BCV
 \end_layout
 
 \end_inset
 
- 
-\begin_inset Flex Glossary Term (pl)
+ (0.417 with GB vs.
+ 0.400 without).
+ The near equality of the 
+\begin_inset Flex Glossary Term
 status open
 
 \begin_layout Plain Layout
-oligo
+BCV
 \end_layout
 
 \end_inset
 
- likely cause very small but non-zero systematic perturbations in measured
- gene expression levels.
+ for both sets indicates that the higher correlations in the GB libraries
+ are most likely a result of the increased yield of useful reads, which
+ reduces the contribution of Poisson counting uncertainty to the overall
+ variance of the 
+\begin_inset Flex Glossary Term
+status open
+
+\begin_layout Plain Layout
+logCPM
+\end_layout
+
+\end_inset
+
+ values 
+\begin_inset CommandInset citation
+LatexCommand cite
+key "McCarthy2012"
+literal "false"
+
+\end_inset
+
+.
+ This improves the precision of expression measurements and more than offsets
+ the negligible increase in 
+\begin_inset Flex Glossary Term
+status open
+
+\begin_layout Plain Layout
+BCV
+\end_layout
+
+\end_inset
+
+.
 \end_layout
 
 \begin_layout Standard
@@ -19435,129 +19549,63 @@ name "fig:gene-abundance-correlations"
 
 \series bold
 Comparison of inter-sample gene abundance correlations with and without
- GB.
-
-\series default
- All libraries were normalized together as described in Figure 2, and genes
- with an average logCPM less than 
-\begin_inset Formula $-1$
-\end_inset
-
- were filtered out.
- Each gene’s logCPM was computed in each library using 
-\begin_inset Flex Code
-status open
-
-\begin_layout Plain Layout
-edgeR
-\end_layout
-
-\end_inset
-
-'s 
-\begin_inset Flex Code
-status open
-
-\begin_layout Plain Layout
-cpm
-\end_layout
-
-\end_inset
-
- function.
- For each pair of biological samples, the Pearson correlation between those
- samples' GB libraries was plotted against the correlation between the same
- samples’ non-GB libraries.
- Each point represents an unique pair of samples.
- The solid gray line shows a quantile-quantile plot of distribution of GB
- correlations vs.
- that of non-GB correlations.
- The thin dashed line is the identity line, provided for reference.
-\end_layout
-
-\end_inset
-
-
-\end_layout
-
-\begin_layout Plain Layout
-
-\end_layout
-
-\end_inset
-
-
-\end_layout
-
-\begin_layout Standard
-\begin_inset Flex TODO Note (inline)
-status open
-
-\begin_layout Plain Layout
-Give these numbers the LaTeX math treatment
-\end_layout
+ GB.
 
+\series default
+ All libraries were normalized together as described in Figure 2, and genes
+ with an average logCPM less than 
+\begin_inset Formula $-1$
 \end_inset
 
-
-\end_layout
-
-\begin_layout Standard
-To evaluate the possibility of 
-\begin_inset Flex Glossary Term
+ were filtered out.
+ Each gene’s logCPM was computed in each library using 
+\begin_inset Flex Code
 status open
 
 \begin_layout Plain Layout
-GB
+edgeR
 \end_layout
 
 \end_inset
 
- causing random perturbations and reducing sample quality, we computed the
- Pearson correlation between 
-\begin_inset Flex Glossary Term
+'s 
+\begin_inset Flex Code
 status open
 
 \begin_layout Plain Layout
-logCPM
+cpm
 \end_layout
 
 \end_inset
 
- values for every pair of samples with and without 
-\begin_inset Flex Glossary Term
-status open
-
-\begin_layout Plain Layout
-GB
+ function.
+ For each pair of biological samples, the Pearson correlation between those
+ samples' GB libraries was plotted against the correlation between the same
+ samples’ non-GB libraries.
+ Each point represents an unique pair of samples.
+ The solid gray line shows a quantile-quantile plot of distribution of GB
+ correlations vs.
+ that of non-GB correlations.
+ The thin dashed line is the identity line, provided for reference.
 \end_layout
 
 \end_inset
 
- and plotted them against each other (Figure 
-\begin_inset CommandInset ref
-LatexCommand ref
-reference "fig:gene-abundance-correlations"
-plural "false"
-caps "false"
-noprefix "false"
+
+\end_layout
 
 \end_inset
 
-).
- The plot indicated that the 
-\begin_inset Flex Glossary Term
-status open
 
-\begin_layout Plain Layout
-GB
 \end_layout
 
-\end_inset
+\begin_layout Subsection
+More differentially expressed genes are detected with globin blocking
+\end_layout
 
- libraries have higher sample-to-sample correlations than the non-GB libraries.
- Parametric and nonparametric tests for differences between the correlations
- with and without 
+\begin_layout Standard
+To compare performance on differential gene expression tests, we took subsets
+ of both the 
 \begin_inset Flex Glossary Term
 status open
 
@@ -19567,32 +19615,35 @@ GB
 
 \end_inset
 
- both confirmed that this difference was highly significant (2-sided paired
- t-test: t = 37.2, df = 665, P ≪ 2.2e-16; 2-sided Wilcoxon sign-rank test:
- V = 2195, P ≪ 2.2e-16).
- Performing the same tests on the Spearman correlations gave the same conclusion
- (t-test: t = 26.8, df = 665, P ≪ 2.2e-16; sign-rank test: V = 8781, P ≪ 2.2e-16).
- The 
-\begin_inset Flex Code
+ and non-GB libraries with exactly one pre-transplant and one post-transplant
+ sample for each animal that had paired samples available for analysis (N=7
+ animals, N=14 samples in each subset).
+ The same test for pre- vs.
+ post-transplant differential gene expression was performed on the same
+ 7 pairs of samples from 
+\begin_inset Flex Glossary Term
 status open
 
 \begin_layout Plain Layout
-edgeR
+GB
 \end_layout
 
 \end_inset
 
- package was used to compute the overall 
+ libraries and non-GB libraries, in each case using an 
 \begin_inset Flex Glossary Term
 status open
 
 \begin_layout Plain Layout
-BCV
+FDR
 \end_layout
 
 \end_inset
 
- for 
+ of 10% as the threshold of significance.
+ Out of 12954 genes that passed the detection threshold in both subsets,
+ 358 were called significantly differentially expressed in the same direction
+ in both sets; 1063 were differentially expressed in the 
 \begin_inset Flex Glossary Term
 status open
 
@@ -19602,7 +19653,8 @@ GB
 
 \end_inset
 
- and non-GB libraries, and found that 
+ set only; 296 were differentially expressed in the non-GB set only; 2 genes
+ were called significantly up in the 
 \begin_inset Flex Glossary Term
 status open
 
@@ -19612,19 +19664,20 @@ GB
 
 \end_inset
 
- resulted in a negligible increase in the 
-\begin_inset Flex Glossary Term
-status open
-
-\begin_layout Plain Layout
-BCV
-\end_layout
+ set but significantly down in the non-GB set; and the remaining 11235 were
+ not called differentially expressed in either set.
+ These data are summarized in Table 
+\begin_inset CommandInset ref
+LatexCommand ref
+reference "tab:Comparison-of-significant"
+plural "false"
+caps "false"
+noprefix "false"
 
 \end_inset
 
- (0.417 with GB vs.
- 0.400 without).
- The near equality of the 
+.
+ The differences in 
 \begin_inset Flex Glossary Term
 status open
 
@@ -19634,44 +19687,31 @@ BCV
 
 \end_inset
 
- for both sets indicates that the higher correlations in the GB libraries
- are most likely a result of the increased yield of useful reads, which
- reduces the contribution of Poisson counting uncertainty to the overall
- variance of the 
-\begin_inset Flex Glossary Term
+ calculated by 
+\begin_inset Flex Code
 status open
 
 \begin_layout Plain Layout
-logCPM
+edgeR
 \end_layout
 
 \end_inset
 
- values 
-\begin_inset CommandInset citation
-LatexCommand cite
-key "McCarthy2012"
-literal "false"
-
+ for these subsets of samples were negligible (
+\begin_inset Formula $\textrm{BCV}=0.302$
 \end_inset
 
-.
- This improves the precision of expression measurements and more than offsets
- the negligible increase in 
+ for 
 \begin_inset Flex Glossary Term
 status open
 
 \begin_layout Plain Layout
-BCV
+GB
 \end_layout
 
 \end_inset
 
-.
-\end_layout
-
-\begin_layout Subsection
-More differentially expressed genes are detected with globin blocking
+ and 0.297 for non-GB).
 \end_layout
 
 \begin_layout Standard
@@ -20103,124 +20143,9 @@ Comparison of significantly differentially expressed genes with and without
 
 \end_layout
 
-\begin_layout Plain Layout
-
-\end_layout
-
-\end_inset
-
-
-\end_layout
-
-\begin_layout Standard
-To compare performance on differential gene expression tests, we took subsets
- of both the 
-\begin_inset Flex Glossary Term
-status open
-
-\begin_layout Plain Layout
-GB
-\end_layout
-
-\end_inset
-
- and non-GB libraries with exactly one pre-transplant and one post-transplant
- sample for each animal that had paired samples available for analysis (N=7
- animals, N=14 samples in each subset).
- The same test for pre- vs.
- post-transplant differential gene expression was performed on the same
- 7 pairs of samples from 
-\begin_inset Flex Glossary Term
-status open
-
-\begin_layout Plain Layout
-GB
-\end_layout
-
-\end_inset
-
- libraries and non-GB libraries, in each case using an 
-\begin_inset Flex Glossary Term
-status open
-
-\begin_layout Plain Layout
-FDR
-\end_layout
-
-\end_inset
-
- of 10% as the threshold of significance.
- Out of 12954 genes that passed the detection threshold in both subsets,
- 358 were called significantly differentially expressed in the same direction
- in both sets; 1063 were differentially expressed in the 
-\begin_inset Flex Glossary Term
-status open
-
-\begin_layout Plain Layout
-GB
-\end_layout
-
-\end_inset
-
- set only; 296 were differentially expressed in the non-GB set only; 2 genes
- were called significantly up in the 
-\begin_inset Flex Glossary Term
-status open
-
-\begin_layout Plain Layout
-GB
-\end_layout
-
-\end_inset
-
- set but significantly down in the non-GB set; and the remaining 11235 were
- not called differentially expressed in either set.
- These data are summarized in Table 
-\begin_inset CommandInset ref
-LatexCommand ref
-reference "tab:Comparison-of-significant"
-plural "false"
-caps "false"
-noprefix "false"
-
-\end_inset
-
-.
- The differences in 
-\begin_inset Flex Glossary Term
-status open
-
-\begin_layout Plain Layout
-BCV
-\end_layout
-
-\end_inset
-
- calculated by 
-\begin_inset Flex Code
-status open
-
-\begin_layout Plain Layout
-edgeR
-\end_layout
-
-\end_inset
-
- for these subsets of samples were negligible (
-\begin_inset Formula $\textrm{BCV}=0.302$
 \end_inset
 
- for 
-\begin_inset Flex Glossary Term
-status open
-
-\begin_layout Plain Layout
-GB
-\end_layout
-
-\end_inset
 
- and 0.297 for non-GB).
 \end_layout
 
 \begin_layout Standard
@@ -20418,7 +20343,7 @@ literal "false"
  The DeepSAGE method involves two different restriction enzymes that purify
  and then tag small fragments of transcripts at specific locations and thus
  significantly reduces the complexity of the transcriptome.
- Therefore, we could not determine how DeepSAGE results would translate
+ Therefore, we could not assume that the DeepSAGE result would translate
  to the common strategy in the field for assaying the entire transcript
  population by whole-transcriptome 
 \begin_inset Formula $3^{\prime}$