P-value cut-off to identify SNPs at ChIP-seq peaks? [BedTools]

Question

[ConvolutedGenome]

Hi all,

I have a bed file of SNPs and also H3K27ac ChIP-seq .broadpeak file from Roadmap epigenome...

I want to find the SNPs in my list that intersects a H3K27ac peak using BedTools intersect

Is the publicly available .broadpeak already completely processed? (meaning I dont need additional filtering) or should I filter the intervals inside my broadpeak file FURTHER based on its pValue or qValue first?? If so what would be the appropriate cut-off for this type of analysis?

Thanks!

Answer 1

[Carlo Yague]

broadpeaks file are already processed, but it doesn't mean that all peaks have the same importance. Filtering can make sense, but it is hard to tell beforehand what is the approriate cutoff for your analysis.

Some questions might help you make your decision: how many peaks do you have ? how many SNPs ? What fraction of the genome is covered by peaks ? If that fraction is small, then I recommend to start by intersecting with the full set of peaks, take a look on the results, then filter if needed.

[ConvolutedGenome]

Thank you for your reply! I'll skip the filtering then since I dont want to be too conservative here

One more question though... (I've looked at previous posts but found no answers) Is it a good idea to merge the available broadpeak and narrowpeak files together? People has suggested that broadpeak is more appropriate than narrowpeak for histone modification, but why would Roadmap provide narrowpeak file as well? Will only using broadpeak lead to loss of information?

[Carlo Yague]

I think its best to use broadpeaks for the reasons you mention. Why would Roadmap provide narrowpeak as well ? Probably because it is an automated pipeline that does both and leave the users decide what is best for their analysis. Generally speaking, broadpeaks and narrowpeaks are different interpretation of the same data, so it should not lead to a real a loss of information to consider one or the other. If you really want to see for yourself the difference between narrowpeak and broadpeak intervals, you can use bedtools intersect on both (instead of merging):

bedtools intersect -wa -wb -a SNPs.bed -b H3K27ac.narrowpeak H3K27ac.broadpeak

:

[ConvolutedGenome]

Thank you, you answered my question :)