how to compare the difference of read counts in a given region

Question

[BenHu]

I have 2 Chipseq, including IP and input samples.

for example, the the number of reads in the chr1:1-500bp is 1000 and 50 in IP and input samples, respectively.
I want to compare the statistic differences for that. Does anyone know how to work on that?

thanks in advance.

Answer 1

[seidel]

The simplest answer would be to use the Poisson distribution. Assuming your data is normalized for read depth, you have a background estimate for your window of interest (lambda = 50), and an observed count for that window in your IP (1000), so in R you could estimate this with:

ppois(100, lambda=50, lower.tail=FALSE)

However, this is only one window, and there maybe many other factors involved. The MACS paper is an easy read which resembles this approach (Zhang et al. 2008, Genome Biology). However, opinions differ on the best distribution to use, as well as other factors involved. See for example: Kuan et al., (2011). A Statistical Framework for the Analysis of ChIP-Seq Data.

[BenHu]

thanks a lot. could you send the titles for both papers. As I can not open the above links.

[seidel]

Sorry about that. I updated the references.