-1
$\begingroup$

What does the padj really mean? I believe it's a BH method, where it attempts to quantify the likelihood of a "false positive" but I am not sure. Also, what does it mean if the resulting calculations nearly all have the same padj values despite having different p-values?

$\endgroup$
1
  • $\begingroup$ It would help to see the data themselves (p-values and adjusted p-values), or at least a general idea of how many hypotheses and the part of the p-value spectrum. For high p-values FDRs ("adjusted p-values", en.wikipedia.org/wiki/False_discovery_rate) will be very similar as they will mostly indicate support for the null hypothesis, i.e. very high false discovery rates. I would suggest reading about DESeq2 stats (section "Multiple test correction"): hbctraining.github.io/DGE_workshop/lessons/… $\endgroup$ Aug 12 at 3:45
2
$\begingroup$

If we look at documentation of DESEQ2 and search "adjusted p-value", we find the section "Multiple test correction". In this section they discuss Benjamini-Hochberg ("BH") false discovery rate (FDR) correction procedures:

FDR/Benjamini-Hochberg: Benjamini and Hochberg (1995) defined the concept of FDR and created an algorithm to control the expected FDR below a specified level given a list of independent p-values. An interpretation of the BH method for controlling the FDR is implemented in DESeq2 in which we rank the genes by p-value, then multiply each ranked p-value by m/rank.

Earlier, they write:

m = total number of tests

They further add:

So what does FDR < 0.05 mean? By setting the FDR cutoff to < 0.05, we’re saying that the proportion of false positives we expect amongst our differentially expressed genes is 5%. For example, if you call 500 genes as differentially expressed with an FDR cutoff of 0.05, you expect 25 of them to be false positives.

I would suggest reading Benjamini and Hochberg (1995) to learn more details if desired.

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.