Difference between padj and FDR

Question

[ML]

I am confuse if padj and FDR is same? I am in the mid of analysis and confused if I have FDR or not. I have seen some explanation but not able to get the answer. please share your views. Thanks

[tothepoint]

Benjamini and Hochberg (BH) defined the concept of FDR and created an algorithm to control the expected FDR below a specified level in a list of independent p.values. As mentioned by @andres.firrincieli DESeq2 adjust p.value by BH algorithm and @i.sudbery clearly explain and @ATpoint already answered the query well on point. padj is FDR.

Answer 1

[i.sudbery]

The "false-discovery rate" is the fraction of positives that are false positives at a given p-value threshold. It is a property of the threshold, not a property of a gene. So, if you do a 1000 tests, and get 100 positives at p<0.05, your FDR is 50% (as 50 false positives would be expected in 1000 tests at p<0.05).

Thus, technically speaking, it doesn't make sense to say a gene has an FDR. This is why many tools will use the term "adjusted p-value" or "q-value". The adjusted p-value is the FDR your experiment would have if you set the threshold at the p-value for this gene. Thus, in our example above, a gene with a p-value of 0.05 would have a padj/qvalue of 50% because if you set the threshold at 0.05, you would have a 50% FDR.

[ML]

Thankyou for this detailed explanation. I also assumed the same but have some doubt so cannot take the risk with the analysis part. Thank you @ATpoint

Answer 2

[ATpoint]

In DESeq2 the FDR-corrected p-values are called padj indeed.

Answer 3

[andres.firrincieli]

FDR is a statistical approach used to adjust your p-values for multiple testing. For example, in DESeq2 the default FDR method applied to your p-values is the Benjamini–Hochberg procedure.