QC of ChIPseq data (controlling false negatives)

Question

[AlexAbdulkaderKheirallah]

Hello All,

I have recently presented some ChIPseq data and one of the criticisms I received was the issue of having sufficient coverage in inferring peaks. I have two biological replicates (different donors) and the data looks very convincing to me due to good inter-donor reproducibility. For the purposes of simplicity I presented the genome-wide distribution of peaks (and their relative heights) on each chromosome and there were some arms of chromosomes that had no observable signal. This lack of signal I saw for each donor separately (independent analyses) and hence my confidence.

Nevertheless my conclusion was questioned on the basis of possible lack of sufficient coverage due to stochastic (random) forces and biases in library preparation. I think it rather unlikely that the same region would appear negative in two independent samples (with independent sequencing) due to stochastic bias. However I want to prove my point and show that I have good (background + ChIP signal) coverage across the genome.

Can you advise me with some software that could calculate coverage and ideally if it can draw some kind of histogram to see how uniform the sequencing across the genome was?

Thanks

Answer 1

[Fidel]

Do you have the matching inputs for the ChIP? Enrichments seen in ChIP/input are more solid compared to just the coverage signal because such coverage is affected by GC, open chromatin and other biases.

For computing coverage you can use deepTools Recently, I added a tool called plotCoverage (this is only in the release-1.6 branch) that makes a histogram like the following:

[AlexAbdulkaderKheirallah]

I do have matching input control. Thanks for your help!

Answer 2

[Ying W]

Have a look at this paper: http://genome.cshlp.org/content/22/9/1813.full

Focus especially on Figure 4G, those metrics are pretty good to use to determine if your ChIP-seq worked or not.