Hi,
In my lab we are processing ChIP-seq all day and we switched to MACS2 beginning this year after using MACS1.4 for some time. We noticed yesterday that MACS2 never call peaks on chromosome X (and I suspect on chr Y too). We played with the p-value / q-value threshold all the way down to 1 (calling everything). We ran the MACS2 with control and without control on the same experiments but were enable to call obvious peaks on the X chromosome. In sharp contrast MACS1.4 call peaks correctly on chrX on the same files. We use BED files as input to MACS as this is our pipeline. BigWig files generated from our BED are looking completely fine.
What I would like to ask the community on is: can you reproduce the same behaviour/problem?
Has this been reported somewhere else? I could not find any thread anywhere.
I contacted the author of MACS2 but no answer yet.
Thanks
I am sure when I used MACS2, I got peaks on chromosome X. May be you could try to visualize the peaks on chrX called by MACS1.4 in your bigwig files created by MACS2.
Thanks for this input. I will look into that.
One more thing - MACS2 filter peaks according to q values by default. Did you use the "-p" argument for p value?
yes. We use p-values as q-values just give either too many peaks or none when there are very few peaks.
With MACS 2, it's recommended by the author that you use q values. I think the idea is q-values are based on empirical data, while p values are based at least partially on the accuracy of the null hypothesis.
Hey, can you comment, how the q-values and FDR calculations linked/correlated in Macs2. Is there a simple conversion, like FDR=10*q.
Thanks
problem solved. MACS2 is working fine. A post-processing step on our side was making an error.