Assessing chIP-seq plotfingerprint IP strength

Question

[tw617]

Hi, I am trying to assess IP strength of some chIP- seq data and am a little confused. My interpretation of this figure: R2 input and Tal1 treatment (green and orange) did not work correctly because the control(input) has the expected normal distribution (almost diagonal line) and the treatment line is similar to this; there is no elbow and the lines are nearly identical so these findings are non-specific. For R1 input and Tal1 treatment, it appears as though the input (control, red) is skewed and the Tal1 treatment (blue) is of a normal distribution (near-diagonal). So does this mean the R1 control (red) had low sequencing depth and the R1 treatment finding is non-specific? There are no sharp rises at high x values so I am thinking that this part of the experiment did not work. What is the correct interpretation of this figure?

Answer 1

[colin.kern]

The R1 input definitely looks like it was not sequenced deeply enough. You can see from the graph that it seems like about 40% of the genome has no reads at all. As for whether you got non-specific binding for the IP libraries, that's harder to determine. The curves you have can be ok for broad histone modifications such as H3K27me3, but since you're doing a transcription factor I would expect more clear enrichment. You can quantify what this graph is showing using the --outQualityMetrics argument of plotFingerprint and giving --JSDsample your input library. You can also go ahead and do peak calling with the R2 data and see if you get an amount of binding sites you expect, and you can check using a program such as Homer whether the binding sites are enriched for the TF motif of your target.