IDR for broad peaks

[rspreafico]

Hi! 

Congrats for this nice piece of software! 

I was wondering whether the latest (2.0.2) version of IDR can be safely used with moderately broad peaks (i.e. something like K4me1, not K9me3). 

Also, would you recommend getting the reproducible peaks just by thresholding on global IDR, or would you recommend running the full pipeline (pseudoreplicates, self-pseudoreplicates, etc) to pick the top max(Nt/Np) peaks?

Thank you for your feedback,

Roberto

[Anshul Kundaje]

Full pipeline is recommended. They all use global IDR. Its just a comparison of the true replicates vs. pseudoreps and selfpseudoreps

-Anshul.

--
You received this message because you are subscribed to the Google Groups "idr-discuss" group.
To unsubscribe from this group and stop receiving emails from it, send an email to idr-discuss...@googlegroups.com.
For more options, visit https://groups.google.com/d/optout.

[rspreafico]

Thank you Anshul! 

And is using the new IDR safe on moderately broad peaks?

Thanks,

Roberto

[Anshul Kundaje]

I cannot guarantee reasonable performance of IDR on broad peaks mainly because peak callers for broad peaks (to put it mildly) really suck. If you did have a stable broad peak caller with a scoring measure that is well behaved IDR would do just fine. So I would say use for broad peaks at your own risk. We have been working on IDR for broad peaks (mostly methods to deal with poor peak calling by getting around the peak caller altogether), but it is still in the test phase. I hope to release this in a month along with new QC methods.

-Anshul

[Ashley S Doane]

Interesting.  

Thank you for this program, it has been very useful.

I was wondering if you have tested IDR with atac-seq data?  Should be similar to DNAse seq, although we find atac-seq signal to be cleaner in terms if background.

I have been using IDR (the new python version) to call reproducible peaks on atac-seq biological replicates.  My input to IDR is a narrowPeak file generated from MACS2, according to methods for atac-seq peak calling in Buenestro et al (Nature, 2015). The narrowPeak file has broad regions, and subpeaks (summits from Macs2), which we use for downstream analysis.  I'm interested in IDR for the broad regions (median peak with ~ 500bp), rather than the subpeaks, so I have been calling -broadPeak for input type, using IDR < 0.1. About 40-60% of peaks are retained after IDR.  This has worked well for my needs.

I'm interested to know if you or others have used IDR for atac-seq, and what guidelines you might suggest.  

Thanks again

-Ashley