Normalization of ChIP-seq results with deepTools

Question

[Daria]

Hello everyone!

I`m trying to visualize ChIP-seq data with deepTools. I used the following commands:

bamCoverage (normalization) -> bigwigCompare (normalization against input) -> computeMatrix -> plotProfile

As a result I got this figure. Green and yellow plots are for treatment samples replicas, blue and dark blue are for negative control replicas.

It seems like I have problems with normalization of samples. The number of peaks in treatment samples is lower than in control, but on the figure the level of the signal in treatment samples is higher for some reason. (Noise does not confuse me. The problem is overall level of the signal) I used parameters "--normalizeUsing RPGC --effectiveGenomeSize 2862010578 --ignoreForNormalization chrX chrY" for bamCoverage command to normalize source .bam files. Apparently, it is wrong and I should use another.

Can anyone help me with this problem and tell the right parameters set? And maybe explain how to choose it correctly?

Thank you!

Answer 1

[ATpoint]

ChIP-seq is a tricky one and simple normalization can be off indeed. Here is my take on how I think it is done properly: ATAC-seq sample normalization

Essentially, you make a count matrix snd then use DESeq2 or edgeR to calculate site factors. You then scale these tracks you use by this factor. Does that make sense to you?