Question: Chip-seq analysis with input
gravatar for damian.kao
2.0 years ago by
damian.kao0 wrote:

I read through the other posts about using RNA-seq packages to analyze Chip-seq data ( It seems like the general consensus is to just ignore the input control samples or build a black-list and look at differential binding between IP samples.

If I do want to incorporate input control into my differential binding, is it valid to include that data as another factor in the design matrix and perform a difference of difference?

So for every library, I would have a "IP" factor with two levels (IP/Input) and also a "sample" factor with two levels for treatment and control. 

I am not very well versed in R, would it be possible to make this kind of contrast? And would this type of contrast even be valid?

ADD COMMENTlink modified 2.0 years ago by Aaron Lun20k • written 2.0 years ago by damian.kao0
gravatar for Aaron Lun
2.0 years ago by
Aaron Lun20k
Cambridge, United Kingdom
Aaron Lun20k wrote:

Can you do it? Yes, because GLMs are very flexible and can easily test for differences of differences.

Should you do it? That's harder to answer, but in most cases, probably not. See A: csaw with negative controls for an explanation of why looking for differences of differences tends to be counterproductive.

ADD COMMENTlink modified 2.0 years ago • written 2.0 years ago by Aaron Lun20k

Thanks. I am trying to understand what you meant when you wrote:

However, this becomes problematic when changes in chromatin state (that affect input coverage) coincide with changes in binding....

Are you saying that for a given genomic region that has signal in the input control, that signal is composed of various factors (mappability, gc content, open chromatin...). One of the factors of input control signal could also be the protein of interest that we are IP-ing. To then try to determine a difference between the protein of interest and the input control (which includes signal from the protein of interest) would not be productive. You would essentially be equalizing out the sample vs input signal.

ADD REPLYlink written 2.0 years ago by damian.kao0

Yes, the risk is that the changes in the input signal (e.g., due to changes in chromatin accessibility) are biologically correlated with genuine changes in ChIP signal (e.g., due to more protein binding in regions of open chromatin). So if you use the former to "correct" the latter, you would weaken or lose the DB effect. Anecdotes suggest that this does indeed happen, which is why I whine about it.

ADD REPLYlink written 2.0 years ago by Aaron Lun20k
Please log in to add an answer.


Use of this site constitutes acceptance of our User Agreement and Privacy Policy.
Powered by Biostar version 2.2.0
Traffic: 212 users visited in the last hour