|
|
@@ -6219,7 +6219,7 @@ name "fig:rulegraph"
|
|
|
|
|
|
|
|
|
\series bold
|
|
|
-Dependency graph of steps in reproducible workflow
|
|
|
+Dependency graph of steps in reproducible workflow.
|
|
|
\end_layout
|
|
|
|
|
|
\end_inset
|
|
|
@@ -6253,21 +6253,119 @@ end{landscape}
|
|
|
|
|
|
\end_layout
|
|
|
|
|
|
-\begin_layout Itemize
|
|
|
-Discuss advantages of developing using a reproducible workflow
|
|
|
+\begin_layout Standard
|
|
|
+The analyses described in this chapter were organized into a reproducible
|
|
|
+ workflow using the Snakemake workflow management system.
|
|
|
+ As shown in Figure
|
|
|
+\begin_inset CommandInset ref
|
|
|
+LatexCommand ref
|
|
|
+reference "fig:rulegraph"
|
|
|
+plural "false"
|
|
|
+caps "false"
|
|
|
+noprefix "false"
|
|
|
+
|
|
|
+\end_inset
|
|
|
+
|
|
|
+, the workflow includes many steps with complex dependencies between them.
|
|
|
+ For example, the step that counts the number of ChIP-seq reads in 500
|
|
|
+\begin_inset space ~
|
|
|
+\end_inset
|
|
|
+
|
|
|
+bp windows in each promoter (the starting point for Figures
|
|
|
+\begin_inset CommandInset ref
|
|
|
+LatexCommand ref
|
|
|
+reference "fig:H3K4me2-neighborhood"
|
|
|
+plural "false"
|
|
|
+caps "false"
|
|
|
+noprefix "false"
|
|
|
+
|
|
|
+\end_inset
|
|
|
+
|
|
|
+,
|
|
|
+\begin_inset CommandInset ref
|
|
|
+LatexCommand ref
|
|
|
+reference "fig:H3K4me3-neighborhood"
|
|
|
+plural "false"
|
|
|
+caps "false"
|
|
|
+noprefix "false"
|
|
|
+
|
|
|
+\end_inset
|
|
|
+
|
|
|
+, and
|
|
|
+\begin_inset CommandInset ref
|
|
|
+LatexCommand ref
|
|
|
+reference "fig:H3K27me3-neighborhood"
|
|
|
+plural "false"
|
|
|
+caps "false"
|
|
|
+noprefix "false"
|
|
|
+
|
|
|
+\end_inset
|
|
|
+
|
|
|
+), named
|
|
|
+\begin_inset Formula $\texttt{chipseq\_count\_tss\_neighborhoods}$
|
|
|
+\end_inset
|
|
|
+
|
|
|
+, depends on the RNA-seq abundance estimates in order to select the most-used
|
|
|
+ TSS for each gene, the aligned ChIP-seq reads, the index for those reads,
|
|
|
+ and the blacklist of regions to be excluded from ChIP-seq analysis.
|
|
|
+ Each step declares its inputs and outputs, and Snakemake uses these to
|
|
|
+ determine the dependencies between steps.
|
|
|
+ Each step is marked as depending on all the steps whose outputs match its
|
|
|
+ inputs, generating the workflow graph in Figure
|
|
|
+\begin_inset CommandInset ref
|
|
|
+LatexCommand ref
|
|
|
+reference "fig:rulegraph"
|
|
|
+plural "false"
|
|
|
+caps "false"
|
|
|
+noprefix "false"
|
|
|
+
|
|
|
+\end_inset
|
|
|
+
|
|
|
+, which Snakemake uses to determine order in which to execute each step
|
|
|
+ so that each step is executed only after all of the steps it depends on
|
|
|
+ have completed, thereby automating the entire workflow from start to finish.
|
|
|
\end_layout
|
|
|
|
|
|
-\begin_deeper
|
|
|
-\begin_layout Itemize
|
|
|
-Decision-making based on trying every option and running the workflow downstream
|
|
|
- to see the effects
|
|
|
+\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
|
|
|
+ that would not have been practical otherwise.
|
|
|
+ For example, 5 different RNA-seq quantification methods were tested against
|
|
|
+ two different reference transcriptome annotations for a total of 10 different
|
|
|
+ quantifications of the same RNA-seq data.
|
|
|
+ These were then compared against each other in the exploratory data analysis
|
|
|
+ step, to determine that the results were not very sensitive to either the
|
|
|
+ choice of quantification method or the choice of annotation.
|
|
|
+ This was possible with a single script for the exploratory data analysis,
|
|
|
+ because Snakemake was able to automate running this script for every combinatio
|
|
|
+n of method and reference.
|
|
|
+ In a similar manner, two different peak calling methods were tested against
|
|
|
+ each other, and in this case it was determined that SICER was unambiguously
|
|
|
+ superior to MACS for all histone marks studied.
|
|
|
+ By enabling these types of comparisons, structuring the analysis as an
|
|
|
+ automated workflow allowed important analysis decisions to be made in a
|
|
|
+ data-driven way, by running every reasonable option through the downstream
|
|
|
+ steps, seeing the consequences of choosing each option, and deciding accordingl
|
|
|
+y.
|
|
|
\end_layout
|
|
|
|
|
|
-\end_deeper
|
|
|
\begin_layout Subsection
|
|
|
Data quality issues limit conclusions
|
|
|
\end_layout
|
|
|
|
|
|
+\begin_layout Standard
|
|
|
+\begin_inset Flex TODO Note (inline)
|
|
|
+status open
|
|
|
+
|
|
|
+\begin_layout Plain Layout
|
|
|
+Is this needed?
|
|
|
+\end_layout
|
|
|
+
|
|
|
+\end_inset
|
|
|
+
|
|
|
+
|
|
|
+\end_layout
|
|
|
+
|
|
|
\begin_layout Chapter
|
|
|
Improving array-based diagnostics for transplant rejection by optimizing
|
|
|
data preprocessing
|
