resume/examples/Salomon/CD4/README.mkdn

<!-- TODO: Update this -->

This is a series of example plots and tables from a combined
RNA-seq/ChIP-seq study on differences between naive and memory T-cell
activation. You can view the (old and messy) code for these plots
[here][1].

[1]: https://github.com/DarwinAwardWinner/cd4-histone-paper-code

- [`p-value distributions.pdf`](p-value distributions.pdf) is a series
  of p-value histograms for each of the contrasts tested. A contrast
  with no significant differential expression would exhibit a uniform
  distribution, while differential expression would be reflected by an
  excess of small p-values.
- [`FPKM by Peak Status H3K4.pdf`](FPKM by Peak Status H3K4.pdf) shows
  the variation in gene expression based on the presence or absence of
  two histone marks in the gene promoters.
- [`promoter-edger-topgenes3-ql.xlsx`](promoter-edger-topgenes3-ql.xlsx)
  is a spreadsheet of all promoters with differential histone
  modification in their promoters based on the ChIP-seq read counts.
- [`Promoter Peak Distance Profile.pdf`](Promoter Peak Distance Profile.pdf)
  shows the distribution of distances from transcription
  start sites to the nearest peak for the three histone modifications
  studied. This was used to determine the "promoter radius" for read
  counting. Notably, the three histone marks do not all have the same
  promoter radius.
- [`rnaseq-MDSPlots.pdf`](rnaseq-MDSPlots.pdf) shows a series of MDS
  plots (similar to PCA plots) before and after correction of a known
  batch effect. Note the implausible zigzag-shaped progression over
  time before correction, compared to the more plausible cyclic time
  progression after.
- [`rnaseq-edgeR-vs-limma.pdf`](rnaseq-edgeR-vs-limma.pdf) and
  [`rnaseq-limma-weighted-vs-uw.pdf`](rnaseq-limma-weighted-vs-uw.pdf)
  show comparisons of p-values for all genes in each contrast of the
  RNA-seq data, comparing edgeR and limma-voom with/without sample
  quality weights. The final choice of method was limma-voom with
  sample quality weights.
- [`rnaseq-maplots-limma-sampleweights.pdf`](rnaseq-maplots-limma-sampleweights.pdf)
  shows the MA plot for each contrast of the RNA-seq data

There are also some plots from an in-progress analysis of the same
data based on sliding windows, rather than just analyzing promoter
regions. You can view the code for generating these plots [here][2],
and you can view some presentation slides based on this analysis
[here][3].

[2]: https://github.com/DarwinAwardWinner/CD4-csaw
[3]: ./ChIP-Seq presentation.pdf

- [`CCF-plots.pdf`](CCF-plots.pdf) shows the cross-correlation
  functions of the ChIP-Seq data for 3 different histone marks, at
  several different levels of smoothing. This plot is used to
  determine the fragment size. You can also observe from the periodic
  wave-like pattern, indicating that multiple adjacent histones tend
  to share the same histone modification.
- [`CCF-plots-noBL.pdf`](CCF-plots-noBL.pdf) shows the same plots as
  above, but without removing reads in so-called "blacklist" regions
  that typically contain high-coverage artifact signals. The result is
  a much messier plot, with many samples having an artifactual peak at
  the read length (dotted line) rather than the actual width of a
  histone (solid line).
- [`site-profile-plots.pdf`](site-profile-plots.pdf) shows plots of
  the relative coverage depth profiles around local coverage maxima in
  the ChIP-Seq data. This plot is used to determine the footprint size
  of the protein being imunoprecipitated. Since this is histone mark
  data, the footprint size should match the size of a nucleosome,
  about 147 bp.
- [`D4659vsD5053_idrplots.pdf`](D4659vsD5053_idrplots.pdf) shows an
  example plot from
  the
  [Irreproducible Discovery Rate](https://sites.google.com/site/anshulkundaje/projects/idr) analysis
  used to identify biologically reproducible peaks in the ChIP-Seq
  data. The plot shows the degree of consistency in the scores for
  overlapping peaks in two biological replicates. Peaks with
  consistently high-ranking scores in both replicates are considered
  reproducible.
- The following reports show QC and exploratory analysis for 3 histone
  marks and
  RNA-seq:
  [H3K4me3](reports/ChIP-seq/H3K4me3-exploration.html),
  [H3K4me2](reports/ChIP-seq/H3K4me2-exploration.html),
  [H3K27me3](reports/ChIP-seq/H3K27me3-exploration.html),
  [RNA-seq](reports/RNA-seq/salmon_hg38.analysisSet_ensembl.85-exploration.html).
  The purpose of these reports is to ensure that the modelling
  assumptions and strategies are appropriate for the data. Sometimes
  several strategies are tested against each other, and the best
  performer is chosen for the subsequent differential
  expression/modification analysis.
- The following reports show the differential expression/modification
  analyses and p-value histograms for the 3 histone marks and
  RNA-seq:
  [H3K4me3](reports/ChIP-seq/H3K4me3-diffmod.html),
  [H3K4me2](reports/ChIP-seq/H3K4me2-diffmod.html),
  [H3K27me3](reports/ChIP-seq/H3K27me3-diffmod.html),
  [RNA-seq](reports/RNA-seq/salmon_hg38.analysisSet_ensembl.85-diffexp.html)
- The RNA-seq data were processed using 10 different combinations of
  quantification pipeline and transcriptome
  reference.
  [`rnaseq-compare.html`](reports/RNA-seq/rnaseq-compare.html) shows a
  series of comparisons designed to investigate the differences
  between these pipelines and references.