Entering edit mode
5.5 years ago
realnewbie ▴ 30
I've used DESeq2 package to analyze DEGs of my RNA-seq data, (I guess) 4 months ago, and now I wanted to double check the results by re-analyzing the data.
Although baseMean values are exactly the same, log2fc, lfcse,stat,pvlaue values have sligtly changed. What could be the reason behind it? Could it be that DESeq2 package is updated and some changes happened on statistics part of it in the past 4 months?
Thanks for your help.
Check if you need to use the lfcShrink function after calling the DESeq2 main function. In older versions DESeq2 would call lfcShrink internally, but in the latest version you have to call it yourself.
This is likely it.
You can look for descriptions of important user relevant changes to R packages in the NEWS file.
Also, here’s a longer post on motivation for the change.
The developers of DESeq/DESeq2 have stated before that slight differences in results between different versions should be expected. In this case, I can only imagine that some transcripts may have been identified as outliers, e.g, by a new Cook's Distance threshold introduced in a later version, or something along those lines.
How much is the change? Could you paste before and after values and provenance information?
ENSDARG00000000001 139.799933351948 0.058625459923395 0.17453323038766 0.335898555210265 0.736947366599281 0.999878824603975
ENSDARG00000000002 532.33641136017 0.065047467376644 0.139490979480544 0.466320242490779 0.640986291495346 0.999878824603975
ENSDARG00000000018 2071.02409512167 -0.044450913332722 0.134081331800732 -0.331522015300261 0.740250214505586 0.999878824603975
"ENSDARG00000000001" 139.799933351948 0.097849753762441 0.29148498450009 0.335693977273847 0.737101647981067 0.999925529820113
"ENSDARG00000000002" 532.33641136017 0.0802371128553763 0.172061759924484 0.466327398316693 0.640981170204464 0.999925529820113
"ENSDARG00000000018" 2071.02409512167 -0.0536265815853227 0.161744261517593 -0.331551679683484 0.740227811448765 0.999925529820113
for that kind of difference, i would like to know seed. Set seed and repeat the analysis, not entire analysis. But the one that generates fc, p and adj p.