Peak Calling In Encode Project

Question

[Vikas Bansal]

Hi all,

I have two questions regarding ENCODE project.

1. Does anyone know, where I could find a list, which tells me the peak calling tool used for each sample (TF, Histones, etc.)? I found a list of peak calling tools (Table S1 in this paper) used in the project but it does not tell me specifically which tool was used for which sample.

2. I searched for the Brg1 peaks for hg18 and hg19 assembly in encode project browser. I downloaded Hela Brg1 peaks for both the assemblies and I was expecting similar number of peaks because only the genome version has been changed. Surprisingly, I found only 65 peaks for hg19 assembly and ~25k peaks for hg18 assembly. I don't understand why there is so much difference in number of peaks since the sample is same and only genome version is different?

Thanks in advance and best wishes,

Vikas

Answer 1

[Vikas Bansal]

I just want to share the reply which I got from the person who is responsible for the Brg1 related queries (2nd point in my original post).

Data from hg18 was converted to hg19 coordinates and scored again using SPP.  (hg18 data was all scored in PeakSeq)  The data should usually look similar, but occasionally one peak caller will be more stringent than the other.  

-Philip

But still, I would also say that from 25,000 peaks with Peakseq (using hg18) to only 65 peaks with SPP (using hg19) is a huge difference.

Answer 2

[Chris Whelan]

I'm not sure about histones, but for TF binding if you start here:

http://genome.ucsc.edu/cgi-bin/hgTrackUi?g=wgEncodeTfBindingSuper

Each of the links to the track sets (HAIB, SYDH, etc.) takes you to a page that describes the tracks and the computational methods used (MACS, etc.)

I don't know about the hg18/hg19 issue - sounds like a bug or technical issue you might want to take up with the UCSC or ENCODE folks?

[Vikas Bansal]

Thanks for your reply. I think your link is not working. But I found a solution for tool used for each sample. If I see the describe table schema in UCSC tables, then for each sample I can find a method description. For second question, may be I will contact ENCODE lab that produced the data.