Metrics for exploring the parameter space of broad domain chip-seq finders?

Question

[Endre Bakken Stovner]

Let us say I have want to use SICER or MACS2 for finding broad domains in ChIP-seq data. These contain plenty of parameters I can tune to change the domains found.

Are there any metrics I can use to see if some parameters perform better than others? Any idea, no matter how seemingly stupid, would be appreciated. If these metrics can be machine inspected so that I can automatically choose parameters that would be preferable, but any suggestions welcome.

Answer 1

[dariober]

This is my answer also to your previous question QC for broad domain chip-seq data?.

At the end of the day, my way of assessing the quality of the ChIP and peak calling is to open ChIP and inputs in IGV and look at a few peaks taken from the high confidence to the lower confidence spectrum. Then check if they "look right". I know, it's subjective, it's not scalable. It sucks.

I was kind of writing a program (https://github.com/dariober/genomeGraphs) to at least simplify the scanning of peaks since scrolling through IGV can be painfully slow.

[Endre Bakken Stovner]

I already starred your repo. Since I use Python, I might go with meta-seq, but great job all the same!