1
$\begingroup$

I have 7 different cell types, 4 normal cell types and 3 disease cell types. I have done until peak calling. Now I want to do peak annotation.

The sample I have is all biological replicates. Should I merge the peak file and then do peak annotation, or should I be doing individual peak annotation?

Improve this question
$\endgroup$

1 Answer 1

Reset to default
1
$\begingroup$

The way we do this (and I think its quite common?), is to merge all the files of the same condition (taking the same number of reads from each), and then calling peaks on the merged sample - so you would have two peak sets - one from normal and one from disease. You then do the union of the two peak sets.

This makes annotation much easier.

For downstream quantitative analysis, you can then quantitate each individual sample against the merged peak set.

Improve this answer
$\endgroup$
8
  • $\begingroup$ since i have already done with the alignment part as you suggested this part "taking the same number of reads from each)," right now i can do but i want to have some preliminary result .So would it be wrong to merge peal files from similar condition and go ahead ? $\endgroup$
    – kcm
    Commented Nov 14, 2019 at 14:49
  • 1
    $\begingroup$ Not "wrong" as such, but you'll get more peaks from the samples that have more reads, which is a bias in the analysis. $\endgroup$
    – Ian Sudbery
    Commented Nov 14, 2019 at 15:04
  • $\begingroup$ I have a doubt , so i do see in paper where they cluster these chip seq or atac seq data.So is it just the tag count data which can be used or it needs some kind of normalisation? just in case of rna seq i use rlog or vst data for clustering or PCA ,which i do get from deseq. $\endgroup$
    – kcm
    Commented Nov 14, 2019 at 19:20
  • $\begingroup$ we've successfully used rlog in the past. $\endgroup$
    – Ian Sudbery
    Commented Nov 14, 2019 at 22:52
  • 1
    $\begingroup$ We take the union peak set as I described above, and then run featureCounts with the files of tags and the union peaks this will give you raw read counts. You need to make sure that featureCounts is configured properly so that it only uses the first base from each read. $\endgroup$
    – Ian Sudbery
    Commented Nov 15, 2019 at 10:34

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.