I used bamCompare (from deepTools) to compute the log2 ratio and see differences in the read coverage between pairs of samples. I have found that some regions the log2 ratio is very high (or low), now I would like to compute a p-values for the log2 ratios to get those that are actually statistically significant higher (or lower), but I am not sure how to proceed. I first thought about assuming a normal distribution of my log2ratio values but after performing some normality tests I've seen that it does not fit a normal distribution (very high peak and very long tails). Does someone know how can I proceed with the computation of the p-value to get the significant log2ratios?

Many thanks!

In essence it's the ratio of two Poisson distributions, but I haven't a clue what that is (and I imagine computing it would be annoying given the presence of zeros). Realistically, you're probably interested in just doing some peak calling.

Edit: Actually, it's probably be more like the ratio of mixed Poisson distributions (I imagine this is an over simplification in practice), since an IP will have background and signal.Thanks for the reply, I guess I'll try another approach than bamCompare.