Search
Question: How to select summit size for histone ChiP-Seq in Diffbind
1
13 months ago by
urjaswita10
urjaswita10 wrote:

Hi folks,

I am new to Chip-Seq analysis hence this may be a basic question. I want to do differential binding analysis using DiffBind on H3K4me3 and HeK4Ac histone peaks called using macs2 in two different conditions. I am following the tutorial and in the count step below they use summits size

tamoxifen <- dba.count(tamoxifen, summits=250)

This example is for TF binding I think but, I am not sure what to use here for my data on histone ChiP-Seq. Could someone please suggest me what to use for summits size for histone ChiP-Seq for DiffBind?

Thanks so much!!

modified 13 months ago by Rory Stark2.6k • written 13 months ago by urjaswita10
2
13 months ago by
Rory Stark2.6k
CRUK, Cambridge, UK
Rory Stark2.6k wrote:

For histone marks that have broad enrichment there is no magic number for this analysis. Are you using "broad" peaks (eg from MACS)?

Something to keep in mind is that the peak width doesn't need to encompass the entire enriched area to do a differential binding analysis. Using the summits parameter in this situation should result in a maximally enriched representative region being used for the comparison. So long as the same interval is used for all samples (which DiffBind will do automatically), the analysis should be reasonable, so it doesn't matter that much so long as the value is comfortably within the actual "peak" widths.

My practical advice here is to try the analysis both ways (with and without the summits parameter). To set the summits parameter, you may want to look at your distribution of peak widths and choose a value accordingly, e.g. the minimum or first quartile value.

For example in the sample dataset:

> data(tamoxifen_peaks)
> summary(tamoxifen$binding[,3]-tamoxifen$binding[,2])
Min. 1st Qu.  Median    Mean 3rd Qu.    Max.
310.0   706.0   830.0   961.4  1073.0  5251.0 

you can see that summits=250 results in intervals of 500bp, which is somewhere between the minimum and the first quartile.

-Rory