Question: GC bias correction deepTools2
gravatar for John
2.8 years ago by
John12k wrote:

Hello all :)

I'm trying to check/correct my ChIP-Seq reads for GC% bias, and I got the following two plots for my Input and for my H3K27ac pulldown. I have excluded repetitive regions, but not mappability or peaks:


enter image description here


enter image description here

So based on the Input plot, I obviously have a little bias which I should correct for going forwards (unless there's a chance I should have called peaks for the input too?).

However on the plot for the ChIP pull down the result is wacky because I haven't excluded my peaks from the analysis. This makes sense since you can't correct for an overall GC bias if your chip is pulling down GC-rich regions specifically - and I understand what to do going forwards, but it seems like a lot of work. Call peaks, exclude regions, normalize the BAM, call peaks again. Is it therefore acceptable to just normalize the ChIP-Seq data to the bias of the Input (since both were sequenced on the same sequencer using the same kit manufacturer)?

All the best :)

deeptools2 deeptools • 1.6k views
ADD COMMENTlink modified 2.8 years ago by Devon Ryan88k • written 2.8 years ago by John12k
gravatar for Devon Ryan
2.8 years ago by
Devon Ryan88k
Freiburg, Germany
Devon Ryan88k wrote:

I can't see the images for some reason, but yes, it's best practice to run computeGCBias on the input and use that with correctGCBias on both the input and the ChIP. Having said that, it's pretty rare these days to have much in the way of GC bias. Who generated the data and when?

ADD COMMENTlink written 2.8 years ago by Devon Ryan88k

Yeah I remember Fidel mentioning that modern PCR kits have a fairer GC specificity, but I have to check these things and this time it looked a little off. This was sequenced on 1st August 2013, so that's quite a while ago. I am currently trying computeGCBias again after excluding all regions with >20 reads, which is a poor-man's-peak-caller, and if that makes my GC plot look flatter, I probably won't bother normalizing. I remember when Fidel first made this tool, some of the PCR kits were really bad -- makes me wonder what would have been the outcome of all the papers published post 2013 if they were re-analyzed these days, hehe.

It's a public holiday in Germany today, so double thank-you for taking the time to help me move forwards on this Devon :)

ADD REPLYlink written 2.8 years ago by John12k
Please log in to add an answer.


Use of this site constitutes acceptance of our User Agreement and Privacy Policy.
Powered by Biostar version 2.3.0
Traffic: 798 users visited in the last hour