advice on DESeq and EdgeR output from bacterial RNA-seq

[ugolino]

Hi there,

I'm analyzing RNA-seq data from a bacterial monoculture exposed to 4 conditions. There are 3 bioreps for each condition and after aligning to reference using bowtie2, getting read counts with htseq (eliminated rRNA), I used both DESeq and edgeR (with TMM) to infer DE or genes. I hope the table below is clear. Column 2 and 3 are DESeq, 4 and 5 edgeR and col 6 is the overlap. FDR with BH correction, FC is fold change. EdgeR detects all the genes that DESeq detects as DE, and appreciably more. I'm really struggling to decide whether to accept only the overlap between the two, or give a chance to the genes additionally detected by edgeR. Recent literature implies slightly higher false pos rate for edgeR at n=3; on the other hand DESeq is sometimes qualified as over-conservative. But I'm no statistician, I'm a biologist and would much appreciate your insight.

thanks!

DESeq edgeR
comparison FDR 5% FC>2 FDR 5% FC>2 overlap
A:B 599 467 896 604 599
A:C 0 0 na
A 550 400 879 573 548
B:C 676 523 972 640 676
B 0 0 na
C 623 443 879 559 622


comparison of RNA-seq methods

http://www.biomedcentral.com/content/pdf/1471-2164-13-484.pdf

http://www.ncbi.nlm.nih.gov/pubmed/22988256

[ugolino]

here is the table again

DESeq edgeR
comparison FDR 5% FC>2 FDR 5% FC>2 overlap
A/B 599 467 896 604 599
A/C 0 0 na
A/D 550 400 879 573 548
B/C 676 523 972 640 676
B/D 0 0 na
C/D 623 443 879 559 622

[ugolino]

Is my post poorly formulated or simply inane? I wonder why is it being ignored?

thanks