Bespoke analyses

Sometimes you may not want to run the DEAGO pipeline in one go. DEAGO can be broken down into four component scripts:

The functionality of all of these scripts is incorporated into the runner script: deago.

The main reason to run the component scripts is when you want to deviate from the standard DEAGO analysis. For example, if you have a two factor analysis where you need to edit the markdown. The examples were designed for the DEAGO tutorial dataset.

Remember to create a new directory for each analysis:

Converting an annotation

Please see Preparing an annotation file for more information on annotation file formats and conversions.

Building a configuration file

If you want to run an analysis with a bespoke markdown file, you must first build a configuration file (see Input files). DEAGO uses this configuration file to set the analysis parameters (e.g. FDR cutoff) and to determine which templates need to be included (e.g. GO analysis).

This command will build a new configuration file using the command line parameters called deago.config which looks like:

annotation_file /lustre/scratch118/infgen/pathdev/vo1/data/ensembl_mm10_deago_formatted.tsv
count_column    7
count_delim \t
count_type  featurecounts
counts_directory    /lustre/scratch118/infgen/pathdev/vo1/data/counts
gene_ids    Geneid
go_analysis 0
go_levels   all
keep_images 0
qc_only 0
qvalue  10E-12
results_directory   /lustre/scratch118/infgen/pathdev/vo1/data/lrt_analysis
skip_lines  1
targets_file    /lustre/scratch118/infgen/pathdev/vo1/data/targets.txt

Build R markdown from templates

The next step in a bespoke analysis is to generate an R markdown file for the standard analysis which can then be edited. It will take the configuration file that you generated earlier and use the values to define the analysis.

This command gives you your R markdown file deago_markdown.Rmd, but does not run the analysis.

Editing your markdown

Now you have a markdown file to work with, you can edit the file to run exactly the type of analysis you want. For example, let's adapt our study design so that it accounts for both cell type and treatment and run a likelihood ratio test (LRT) (see the DESeq2 vignette for more information on study design and hypothesis testing).

You can use a command line editor (e.g. nano) to edit the markdown file:

First load the DESeq2 library in your markdown (around line 29) as you will be using functions from the package.

library(DESeq2)

Now let's edit the study design (this will be around line 60) replacing:

dds <- runDESeqAnalysis(targets, countdata, parameters)

With the following code:

coldata <- DataFrame( cell_type=factor( tolower(targets$cell_type) ),
                      treatment=factor( tolower(targets$treatment) ),
                      condition=factor( tolower(targets$condition) ) )

dds <- DESeqDataSetFromMatrix(countData=countdata, colData=coldata, 
                              design=~cell_type+treatment+cell_type:treatment)

dds$cell_type <- relevel(dds$cell_type, ref = "ko")
dds$treatment <- relevel(dds$treatment, ref = "ctrl")

dds <- DESeq(dds, test="LRT", reduced=~cell_type+treatment)
dds <- annotateDataset(dds, parameters)

We can now remove the existing contrast tests, leaving the R session at the end of the analysis and then modify the contrasts to reflect our LRT results:

lrt_results <- results(dds, alpha=as.numeric(parameters$qvalue))
lrt_results$symbol <- mcols(dds)$symbol

contrasts <- list('lrt'=lrt_results)

Finally, replace all occurences of the contrast name ko_il22_vs_ko_ctrl with lrt.

Take a look at the configuration file, modified R markdown file and HTML report that were generated or see the full inputs and results here.

Converting markdown to HTML

Now you just need to convert your modified markdown into a HTML report.

This should give you your HTML analysis report: deago_markdown.html.

Return to the index
Previous: GO term enrichment