search

jsh58 / MACS

MACS -- Model-based Analysis of ChIP-Seq
http://liulab.dfci.harvard.edu/MACS/
Other
6 stars 2 forks source link

DNase-seq analysis #1

Open jsh58 opened 5 years ago

jsh58 commented 5 years ago

(I received this question by email, and am including it and the response below - jmg)

I read your work on improved peak calling with MACS2. I belong in the worst category and have been running MACS2 with -f BAM :) Hence, I'm trying it out as you suggest on a DNaseI data set of my own. Out of interest, would you know if each read pair is treated as one or two events when running with -f BAMPE?

I would argue that it should be two, one for each fragment end (=DNaseI cut), but the way you write it in the paper, I understand that the two reads are used to resolve the original fragment ==> one event.

If you could help me out, it would be much appreciated.

jsh58 commented 5 years ago

It is certainly a good question, and my best answer is that neither of the modes of MACS2 is optimal for DNase-seq. DNase-seq is similar to ATAC-seq, so I will refer you to the section "Interpretation of alignments suitable for ATAC-seq" in the ATAC-seq guidelines, Figure 4 and the following paragraph. Consider using Genrich, it's good and fast.