proper normalization of ChIP-Seq data

[HESmith]

We're comparing a histone modification in control samples vs. experimental samples in which the modifying enzyme is overexpressed. Western blotting indicates ~3X increase in the amount of modified histones, and qPCR of ChIPed material shows a similar increase for known target sites. Comparison of ChIP-Seq data sets indicate that the distribution of the modification is unchanged (i.e., the same peaks are detected in control and experiment). This result was anticipated by other experiments, so nothing earth-shattering.

The question: what is the appropriate method/algorithm for normalizing the ChIP-Seq data that demonstrates the enrichment we observe at these identical peaks? Most of the available tools seem designed for detecting different peaks, using the same (normalized) number of reads. In our case, however, such normalization essentially suppresses the difference in magnitude we detect by the other methods.

I'd appreciate suggestions from the more statistically-savvy members of the community.

Thanks,
Harold

[aurélie teissandier]

Hi Harold,

I have the same problem. Do you have an answer to your question? Which method do you use to compare your different samples?

Thanks
Aurélie

[mknut]

To investigate differences between peaks at the same positions I use DiffBind. It looks like you have already done peak occupancy analysis, the next step would be binding affinity analysis. Binding affinity analysis is basically investigation of differences in enrichment at same peaks across different samples - seems to be exactly what you are after. DiffBind is an R package and is actually quite easy to use and does not require a lot of expertise in R.