Comparing ChIP-seq data sets for different mutants

Hi,

I have a fairly trivial beginners' question: I'm wondering about the best way to visualize differences in histone ChIP-seq data for different mutants in IGV....

I have K4me3 ChIP-seq for WT and three mutants (2 biological replicates each). In order to visualize smaller differences between the different samples I want to calculate Mutant/WT difference data.

My questions: Best to normalize each sample to Input first, then calculate log2 ratios for mutant-WT (I used to this for microarrays in the past) OR can I just calculate log2 ratios for the IPs for mutant-WT (without Input normalization first)? Would you calculate ratios for biological replicates separately or combine them first? What would be the best tool to use? So far, we have used deeptools bamCompare.

Many thanks, Mischa

chip-seq igv normalization

• 938 views • link

updated 5.5 years ago by ATpoint

• written 5.5 years ago by Mischa •

1 answer

I like to create some bigwig files first for every IP, normalize it e.g. A: ATAC-seq sample normalization (quantil normalization) and then compare these files, e.g. via profile plots. I personally do not do input normalization, mainly because there is not really a robust method for this out there that is well-accepted, at least I am not aware of one (not meaning that it does not exist). I would not plot log-ratios but rather the normalized counts directly in a profile plots as with ratios you can see large differences when the counts are low, but these differences are typically unreliable. Imagine like 10 counts vs 1 counts, a 10x change, sounds impressive, but if the error margin of the assay is like +/- 10 counts then this can easily change into a 0x change or a 20x change, in any case not reliable. If you deal with ratios you need statistics to back it up, e.g. the csaw package of DiffBind come to mind for pairwise comparisons of peak profiles.

• 0 views • link

written 5.5 years ago by ATpoint