Normalizing chip-seq signals with different number of peaks

Question

[Rajendra KC]

Hello all,

I want to compare chip-seq signals of polymerase1 and polymerase3 over few bed coordinates but I guess I must be normalizing these chip-seq data before doing any types of quantitative analysis. Polymerase1 chip-seq has lower signal to noise ratio. Further, polymerase1 has considerably less binding sites(signal peaks) in genome than polymerase3. How can I normalize these chip-seq data?

Answer 1

[jared.andrews07]

Direct quantitative comparison of different TFs, polymerases, or histone marks doesn't really work. As you noted, the antibodies will have different efficiencies and as such, will not be directly comparable.

You can, however, look at them more qualitatively, such as taking the pol1 peaks and viewing signal for pol3 in those same regions (and vice versa) as can be done with plotHeatmap or plotProfile from deepTools. Or looking at correlation between signal between the BAMs, which helps ignore general signature magnitude differences. multiBamSummary and plotCorrelation will allow you to do this.

These methods will make more sense in your case and help you to identify regions that are (dis)similar between the two polymerases.