Kolmogorov-Smirnov as a way of comparing methylomes

[Ge0rges]

Hi @ArtRand,

I wanted to mention here a statistical test that might prove a useful contribution to modkit. In my study I have been trying to ask the question "How different are two methylomes?". In the last few weeks, I have settled on the Kolmogorov-Smirnov test which has proven quite useful. It gives you a p-value of significance as to whether the distributions are identical or not, and when significant it gives you the d-value indicating broadly the largest distance between the distributions on an understandable scale of 0 to 1.

My implementation handles the large number of methylation sites by basically bootstrapping the KS test, then calculating the p-value from the number of times the resulting d-value is above the observed d-value. I then do a multiple test correction as well on the p-value.

Let me know what you think!