ChIP-seq normalization method when global loss of signal is expected.
2
0
Entering edit mode
10 months ago
crimacar • 0

Hi. I am studying a protein that potentially affects H2A.Z levels by regulating the complex depositing this histone variant (SWR1c). So we did ChIP-seq in WT, SWR1c knockout and a knockout of my protein. We sequenced input, H2A.Z and H2B. I would like to plot heatmaps and metagene plots, but I have no idea what would be the best normalization to use.

1) Lots of posts don't recommend the use of FPKM or TPM. It seems that TMM, RLE or quantile normalization are generally better methods for ChIP normalization. However, these methods are not a good option when there's loss of signal at a global level, as it happens for H2a.Z in the SWR1 knockout. Which method would be more appropriate?

2) We don't expect global loss of H2A.Z when our protein is knocked out, only a subset of genes affected. So, in order to compare between samples, should I use the same normalization method in both, even when it's not de most appropriate for one of them? Or should I normalize each one with a suitable method?

3)In addition to that, one of my libraries failed and I have one sample with half the reads than the others, so I should normailze to correct that too.

Do you have any suggestions about the best strategy?

Thank you very much for your help.

ChIP-Seq normalization • 510 views
ADD COMMENT
0
Entering edit mode
10 months ago

This is one occasion when spike-in chromatin may have saved you a lot of hassle. Regardless, there are a few methods to deal with global histone changes, but the one I am most familiar with is called ChIPseqSpikeInFree. You might try giving that a shot.

ADD COMMENT
0
Entering edit mode

Thanks for your suggestion! I've just tried ChIPseqSpikeInFree, but I am not sure if I understand it correctly. You are supposed to get larger scaling factors for lower histone mark levels, but I got SF=1 for all my inputs and knockouts, and SF=1.44 for WT. I ran my data through SICER and it detects around 10000 peaks with decreased histone levels (even without spike-in), so I'm sure there is less histone in my mutant sample...

ADD REPLY
0
Entering edit mode

Have you confirmed the global changes via western blot or similar? That is going to be the easiest way to determine if the effect is real or if the IP for the knockout in question just didn't work as well.

ADD REPLY
0
Entering edit mode
10 months ago
Rory Stark ★ 1.2k

There are good discussions of these issues, with examples, in the vignettes for the Bioconductor packages DiffBindand csaw:

DiffBind Vignette

csaw User's Guide

ADD COMMENT

Login before adding your answer.

Traffic: 2313 users visited in the last hour
Help About
FAQ
Access RSS
API
Stats

Use of this site constitutes acceptance of our User Agreement and Privacy Policy.

Powered by the version 2.3.6