Hello everyone,
I have three replicates for my ChIP-seq data, i did peak calling using MACS2, now i was thinking to get the common peaks from all the replicates which in my sense can be called as significance peaks. Now i have some questions 1. Which tool or algorithm should i use to do that and also i used "bedtools intersect" for getting overlapping peaks but the problem is after getting final bed file with common peaks i'm getting some redundancy in the file as intersection was done as per coordinates of the peaks and also for motif analysis which summit position should be used as after intersection it is very tough to understand. 2. As all the files has different summit position which summit positions should be used for motif analysis.
Thanx
•
written
by
chaudharyc61 •
bedtools intersect would indeed be a reasonable choice, e.g.
bedtools intersect -a 1.narrowpeak -b 2.narrowpeak -u | bedtools intersect -a - -b 3.narrowpeak -u | cut -f1-3 > reproducible.bed
If you want summits (given that the center position of reproducible.bed does not serve as a sufficient proxy) you can merge all three replicates at more or less equal contribution (equal number of reads each) and call peaks on this. Then intersect this peak file with reproducible.bed to get summits that more or less represent the "average" in the dataset.
Keep in mind tough that this is a pretty stringent way of filtering. Maybe one of the samples is of poor quality so 3-way intersection would massively reduce the number of peaks in the output even tough two samples might be totally fine. Be sure to only use good-quality samples.
The results of 'bedtools intersect -u' change depending on the order of the samples, which I think is unwanted if we want to generate consensus peaks.
bedtools intersect -a 1.narrowpeak -b 2.narrowpeak -u | bedtools intersect -a - -b 3.narrowpeak -u
Should give slightly different results from
bedtools intersect -a 2.narrowpeak -b 1.narrowpeak -u | bedtools intersect -a - -b 3.narrowpeak -u
Log in to answer this question.
I would recomment to use a "consenus" peak set which contains peaks found at least in two of your replicates. Here is a description of how to: https://ro-che.info/articles/2018-07-11-chip-seq-consensus