batch correction in ChIp-seq data

Question

[Bogdan]

Dear all,

when working with a set of ChIP-seq data that is collected at multiple time points, which algorithm would you use (if any) for batch effect correction ? or for ATAC-seq ? Thanks,

Bogdan

[rpolicastro]

Are you talking about batch correction for something like differential peak calling? Programs like diffbind use DESeq2 or edgeR under the hood, so you can correct for batch in the regression formula the same way you would for RNA-seq. This is assuming you have the correct experimental setup to correct for batch.

[Bogdan]

Thanks a lot. A quick question please just to confirm : for batch correction of ChIP-seq, would we need at least 2 replicates per condition, correct ?

[rpolicastro]

For differential peak calling you want at least 2 biological replicates, and to correct for batch via regression covariate you want samples of each condition per batch. Here would be an example experimental setup.

   condtion batch
1 Untreated     1
2 Untreated     2
3   Treated     1
4   Treated     2

Before correcting for batch check to see if samples are separating by batch in the first few PCs.

[Bogdan]

great, thanks a lot !