Enrichment Analysis from DESeq2 data

Question

[gkuffel22]

Hi everyone,

I use DESeq2 for all of my RNA-Seq analysis. Can anyone recommend a tool that works seamlessly with DESeq2 data to perform enrichment analysis? Thanks everyone.

[WouterDeCoster]

Hello gkuffel22!

It appears that your post has been cross-posted to another site: http://seqanswers.com/forums/showthread.php?p=209714

This is typically not recommended as it runs the risk of annoying people in both communities.

Answer 1

[arnstrm]

If you have a list of genes that are deferentially expressed, you can run enrichment analysis (not just DESeq results). I use ontologizer, which is a command line tool and very easy to run. Here is how to run:

1. Before you begin, you need to have 2 important files. 1. GO ontology file, where you describe every possible GO term numbers with what they are. You can easily get this from here: http://purl.obolibrary.org/obo/go.obo 2. Mapping file, this will describe the association of genes of the organism with GO terms (also available on GO website, if it is a well known organism, else you may have to generate your own).

2. Once you have these 2 files, you can run Ontologizer

   java -jar Ontologizer.jar -a association.anno -g gene_ontology.obo -s your_input_list.txt -p population.txt -c Parent-Child-Union -m Westfall-Young-Single-Step -d 0.05 -r 1000
   

   Here the population.txt is basically the full list of genes that you have in anno file. The only time you need to change this is when you have a different background set to test. You can also play around the other settings like --mode, --resamplingsteps to optimize. With the -d it will also generate a dot file, that can be used with GraphViz to show the pathway where these genes are enriched.

3. To generate graphs:

   dot -Tpng input.dot -o output.png
   

Hope this helps! (you can also run BiNGO, a cytoscape plugin which is easier if you are new to command line!)

Answer 2

[moxu]

I use GSEA (Broad Institute). Just rank your genes by -log(P-value) * sign(logFC).

[langya]

why use this value instead of log2FC?

[moxu]

log2FC is not stable, lower p-value is more stable. I would weigh true signal more than big signal. My personal experience also supports this view. Actually, ranking by -log(P-value) * sign(logFC) is identical to ranking by signal-to-noise ratio.

Answer 3

[EagleEye]

GeneSCF for your significantly differentially expressed genes (gene symbols or ids) from DESeq2.