Question

DESeq2 Results in Different Pvalue & Log2FC using the same code and data

0

Entering edit mode

Xu • 0

@e1d455e5

Last seen 10 weeks ago

United States

Hi there, I am a bioinformatics student at UT Austin. I recently ran our lab's code from 2020 using DESeq2 on an RNA-seq dataset. The output from my recent DESeq result has slightly different log2FC and Pvalue in most genes (roughly 10^-5 or smaller) when compared with the 2020 DESeq result. I am sure it's the same code and RNA-seq data as the 2020 result.

Has there been any update of the DESeq2 package over the last several years that may result in this difference in results?

Thanks.

Best, Xu

DESeq2 dese • 246 views

ADD COMMENT • link updated 10 weeks ago by ATpoint ★ 4.0k • written 10 weeks ago by Xu • 0

0

Entering edit mode

By the way, my 2024 DESeq result and 2020 DESeq result have the same baseMean, which should indicate that my original RNA-seq data was exactly the same for both results.

I checked this by using identical(DESeq_result_2020[, 'baseMean'], DESeq_result_2024[, 'baseMean'])

ADD REPLY • link 10 weeks ago Xu • 0

score 0 · Answer 1 · 2024-02-14

Generally, exact reproducibility is only guaranteed if you use the same exact software version. Many updates happened to the package, see https://bioconductor.org/packages/release/bioc/html/DESeq2.html under "NEWS". The effect seems minor, so either go with the old results for exact consistency, or simply use the latest version which is usually a good idea since it is the most advanced state of the art.