RPGC normalisation creates artefacts at centromere

Question

I am studying ChIP-Seq data in HeLa cells and I've started using the RPGC normalisation of deepTool's bamCoverage. MACS2 also uses this normalisation for its peak calling.

I am seeing a large number of peaks (like, over 50% of my peaks) at the centromere of each chromosome, and the RPGC normalised BW files also show this increased signal at the centromere.

Does this mean that my protein really is at the centromere of each chromosome, or is it some normalisation bias I'm not aware of?

Here's the general aspect of my peak calling (in green). Notice the large increase at the centromere of each chromosome.

And here's a zoom on the centromere of chromosome 6. In blue at the top is raw BW, then there's a lane for input normalised BW but not RPGC, then green is the peak calling by MACS2, and black is RPGC normalised BW.

Answer 1

UPDATE

Here is some other protein ChIPs that I ran through the same normalisation:

As you can see, they don't seem to be enriched at the centromere. So unless anyone brings other contributions, I will accept this as answer.

RPGC normalisation doesn't appear to create artefacts at centromere. The peaks observed here are as real as can be.