Recommendation for using limma-voom workflow for ChIPseq analysis. Or other tool for capturing global histone mark difference?

Question

[m.sadman.sakib]

Hello awesome community!

I just wanted to know your expert opinion, whether limma-voom protocol is suitable for analysing ChIPseq data(with macs2 peaks and raw read counts generated from featureCounts). Does it assume that majority of the expression will not change among groups, just like DESeq2? I am having trouble with my ChIPseq data analysis as I am expecting global change of histone modification, which doesn't follow DESeq2's assumption. DiffBind seemed to be working to some extent to capture the global change with its simple BAM file library size normalization, but my experimental design is too complex to block multiple unwanted factors and get proper differential binding regions.... I was thinking of bypassing DESeq2's default sizeFactor normalization and feed simple BAM library size factors(just like DiffBind) and then doing estimateDispersions and nbinomWaldTest functions to perform differential analysis. I know in this case, running a spike-in control would have been optimal, but it is not possible right now.

I would really appreciate if you can comment on which sets of tool should be optimal to capture global histone modification differences when spike in control is not available.

[ATpoint]

Check the csaw package. It manual contains extensive suggestions for analysis of ChIP-seq data.

[m.sadman.sakib]

Thank you! I will look into it. My current knowledge tells me, it is a window based tool, rather then "analysing predefined peaks(by macs2, for example)" that I am doing. Could you comment if csaw is more advantageous than defined peak based analysis?