QC for broad domain chip-seq data?

Question

[Endre Bakken Stovner]

If you do narrow peak calling you can do autocorrelation as a form of QC. This is just comparing the peaks from the two-strands for the same potential peak.

I cannot think of anything similar that would work on broad domains.

Some stupid ideas for QC to get the creative juices flowing:

• Find regions that are always eu/heterochromatic in all tissues according to the epigen roadmap and see that those respective regions contain H3K4me3/H3K27me3 respectively in the data you are analysing...
• Investigate genes you know are supposed to be on/off in the data you have to see whether they are considered enriched for the appropriate histone modification...

Can you think of anything else?

[Devon Ryan]

You can still look at correlation, just don't do it with peaks (see, for example, multiBamSummary and plotCorrelation in deepTools.