How to calculate enrichment significance for concrete genomic regions?

[ge2sasag]

Hi all,

I´m newbie in chip-seq data analysis

I have two different chip-seq data (one is mine and the other from ENCODE) and I would like to calculate the statistical significance for the enrichment of these two datasets in some concrete genomic intervals in which I´m interested in. I don´t know which kind of analysis I should perform for this. What I want to check is if the enrichment that I see is really significant or it can be just by chance.

Hope you can help!
Big thanks!
/Gema

[dpryan]

The general idea is to use a hypergeometric test (namely fisher's exact test). Google around for that or "contingency table" for details. This happens to be the most frequently used approach for things termed "enrichment analysis".

[ge2sasag]

Thank you for your suggestion!

[ge2sasag]

I have one more question, to run the fisher´s exact test, which input file should I use? sam, bam, wig?

[dpryan]

Number of reads/fragments mapping to the regions of interest in each group and number of reads mapping elsewhere in each group. I'm assuming from your original question that you want to see if your data is different from that from Encode.

[ge2sasag]

yes, I´m analyzing some putative enhancer regions defined by H3K27ac (this dataset is from Encode) and I observed that my transcription factor of interest is also enriched in those regions but I want to know if my factor enrichment is significant or is just by chance because the enhancer regions are big.