How to select top significant ChiPseq peaks from MACS2 output file

Question

[Mike]

Hi all,

I have Macs2 output results in xls format , there are 10 columns in results files (including -log10(pvalue), fold_enrichment & -log10(qvalue), How to select top significant peaks from this file.

Thanks in advance

Answer 1

[igor]

Since p- and q- values are measures of significance, you can sort by those values. Since they are -log10, the most significant are the one with the highest values.

[Mike]

Great, Thanks a lot,

And what about fold_enrichment ?

Answer 2

[dariober]

In practice I would suggest to look at some peaks with different fold changes to get a feel for what a sensible threshold should be for a peak to be convincing. Often I find that peaks with fold change ~2 don't look "peaky" at all so I discard anything below that threshold. Then for ranking peaks, it shouldn't matter very much if you use fold change, p-value or qvalue since these are very well correlated.

Also, take care that peaks with very high fold change (or very low p/q value) can be suspicious.

[igor]

True. You can also use something like IDR to refine your peak list.

But it really depends on what your goal is. I've seen papers where they select the top 500,000 peaks. Clearly, at those numbers, many of them are not going to be very good at all.