How to make complex contrasts in DiffBind dba.contrast

Question

[Bioinformatist Newbie]

Dear All,

I am analyzing ChIP-seq data where I have multiple groups. I am interested in getting differential binding across complex contrasts like:

((condition1 vs. condition2) vs. (condition3 vs. condition4))

I am unable to figure out how to provide such complex contrast to the dba.contrast function in the DiffBind R package. If somebody has worked on such a study kindly share your solution with me.

Thanking you in advance!

Note: That's how my condition column looks like:

metadata$Condition
 [1] 1 1 1 1 1 1 1 2 2 2 2 2 2 2 3 3 3 3 3 3 3 4 4 4 4 4 4 4
Levels: 1 2 3 4

and that's how I am calling the dba.contrast function:

DAPs <- dba.contrast(DAPs, categories=DBA_CONDITION)

As per my understanding by following this rationale, I can get simpler contrast e.g. condition1 - condition2 by the call like:

DAPs.DB <- dba.report(DAPs, 1, method=DBA_EDGER, DataType=DBA_DATA_FRAME, th=0.05, bUsePval=TRUE, bCalled=TRUE)

or condition3 - condition4 by the call like:

DAPs.DB <- dba.report(DAPs, 2, method=DBA_EDGER, DataType=DBA_DATA_FRAME, th=0.05, bUsePval=TRUE, bCalled=TRUE)

Answer 1

[Rory Stark]

In the current version, DiffBind does not support complex model designs and contrasts. However in the next release (currently in the development version), arbitrary model designs and complex contrasts are supported.

Ir order to get the contrast you want, you'll need to reconfigure your metadata a bit and use two factors to pick out all the groups.The edgeR vignette has some examples similar to yours.