Does sequencing depth affect MACS2 q-values?

Question

[ivoryec]

Does sequencing depth affect MACS2 q-values?

I have input A and IP A, and I used MACS2 to call peaks. I also have input B and IP B, and the B set of peaks.

Set A had about 3x more sequencing depth than set B.

The two sets of peaks have ~similar~ distributions of fold enrichment values but the -log(q-value)s are generally much higher in set A.

Should I attribute the difference in q-values to the difference in sequencing depth? If not, how else might I explain this difference in q-values?

Thanks!

Answer 1

[Ian]

There used to be (v1.4) a diagnosis option (--diag) that tested what happens when a random 10%, 20%, 30%, etc. of the reads were used in the analysis. But this does not appear to be available in MACS2. You could do this yourself by selecting sub-sets of reads, and repeat for the replicate. The qvalue is dependent on the distribution of P-values within an experiment, so it is possible if the distribution of P-values differs between replicates. You could try plotting the distribution of Pvalues for yourself.