Strange Result in Deseq Output

[cburke11]

I have a quick question about Deseq. I have 4 conditions in which there is a strain factor and a day factor (eg. Day 1 Strain 1, Day 2 Strain 1, Day 1 Strain 2, Day 2 Strain 2). I have all of these in a large data matrix in Deseq with all conditions and their replicates (n=2 for each point).


row.names DBE DUZ D7UZ D7BE DBE2 DUZ2 D7BE2 D7UZ2
1 CL10000Contig1 20 222 75 50 8 19 206 183
2 CL10001Contig1 1 1 74 257 45 20 174 10
3 CL10002Contig1 67 222 316 441 87 95 561 320
4 CL10003Contig1 1 39 69 61 21 10 377 36
5 CL10004Contig1 10 68 203 232 54 127 249 156
7 CL10005Contig1 0 0 0 0 26 0 0 0

For some of the rows in this matrix this means that expression level may be at 0 for a locus for all data points except for 1 condition. I am next doing pairwise comparisons between 2 conditions at once. My problem is that in my out put I have a list of significant genes returned by "resSigD <-resD[resD$padj <0.1,]" but at the bottom of this list I have some row that are listed as NA.1, NA.2, NA.3 and so on for about 400 rows.. do you know what is causing this problem? Could it be that I am making a pairwise comparison between two conditions but all expression values are 0?

Any insight would be greatly appreciated!

Thanks! Please let me know if you need more info

[Simon Anders]

If a gene has zero counts in all samples involved in the comparison, the p value is NA. This confuses the subsetting. Use "resSigD <-resD[which(resD$padj <0.1),]" to get rid of the extra NAs.

[cburke11]

Thank you for your help Simon. That seems to do the trick