View Single Post
Old 06-27-2013, 12:31 PM   #1
Location: maryland, usa

Join Date: Oct 2011
Posts: 14
Default DESeq: intepreting results from interaction analysis

Hi there,

I am using DESeq to infer differential gene expression in a 2-factor design. Factor 1 (strain) and Factor 2 (condition) each have 4 levels; 3 bioreps each totaling to 48 samples. I am interesed in identifying an interaction effect between strain and condition, on gene expression:

> fit1 = fitNbinomGLMs( cdsFull, count ~ strain + condition + strain : condition )
> fit0 = fitNbinomGLMs( cdsFull, count ~ strain + condition )
> pvalsGLM = nbinomGLMTest( fit1, fit0 )
> padjGLM = p.adjust( pvalsGLM, method='BH' )
> DEresults <- transform( fit1, pval=pvalsGLM, padj=padjGLM )

Attached are results for few significant genes. I am trying to understand the fold changes and p values.
1. How do I identify which strain and which condition have a significant interaction, ie which comparison is the padj referring to? Do I need to do pairwise comparisons of each strain against each condition?

2. If the Intercept represents foldchange ratio of strain1 : condition1, are the other columns subtracted differences in fold change?

I can't seem to grasp the output, any tip is greatly appreciated.

Attached Files
File Type: txt interactionSample.txt (1.7 KB, 16 views)
ugolino is offline   Reply With Quote