DESeq2 inconsistent results

Question

[Chris Gene]

I'm getting inconsistent numbers of genes from DESeq2 results.

Using the default FDR,

On summary(res), I have:

out of 26572 with nonzero total read count
adjusted p-value < 0.1
LFC > 0 (up)     : 2117, 8% 
LFC < 0 (down)   : 1675, 6.3% 
(i.e., total `3792`)

However, if I subset using either of the following, I get 2 less genes (3790):

resSig <- res[which(res$padj<0.1),] 
resSig <- subset(res, res$padj<0.1)

This really doesn't make any sense at all! Any thoughts?

I also get a different value for padj<0.05 if I subset the above results table or if I run DESeq with alpha=0.05, thoughthat is answered by Michael Love here: http://seqanswers.com/forums/archive/index.php/t-41887.html

[GouthamAtla]

I guess you should also take care of NA

(!is.na((res)$pvalue)) & (!is.na(res$padj))

[Carlo Yague]

have you tried with "less or equal" instead of just "<0.1" ?

[Michael]

Please provide the output of:

sessionInfo()
sum(res$padj=<0.1, na.rm=T )

sessionInfo is always required otherwise we don't know which version we are talking.

[Chris Gene]

When I reopened the session to get the session info as requeste, the error did not reproduce... I happened to update DESeq2 in the interim - i wonder whether that was it...