annotation of chipseq peaks

Question

[rob.costa1234]

I have a Chipseq data with peak coordinates that I want to overlay with 3 public data sets associated with the same protein. I have used bed tools to intersect coordinates to check the overlap of my data with public datasets. I was wondering if it is reasonable do the Go/ pathway analysis with peaks of public datasets and overlapping peaks of my dataset and look for common biological terms. Since the definition of gene may be different so I can assign common distance in both datasets (public and mine) and will provide more meaningful biological insight. On the other side since peaks are overlapping (not exactly but with few bases) it can be argued that both peaks (public and my dataset) will have the same pathways. I can use GREAT for this purpose. Any suggestion will that be a reasonable approach to look for consensus Go terms or not? BTW I am not asking for my assignment. I thought about it and googled it but did not find any paper showing what I am thinking. So I am requesting expert feedback from the community.

[Carlo Yague]

In my opinion, "consensus GO terms" as you propose is not the most meaningful way to look at this data. Instead, I would

1. define the "consensus peaks" across datasets -> what is the proportion that overlap (you could make a Venn diagram for instance) ?
2. perform GO term enrichment analysis on the consensus peaks.

Answer 1

[Qiongyi]

Regarding the overlay analysis with multiple data sets, you can use multiIntersectBed (in the bedtools package). Venn diagram is a good way to present the overlay results as Carlo suggested.

In general, what you proposed sounds reasonable. Not sure how many peaks (%) can be assigned with GO terms as most peaks may be not sitting in a gene region. If there is a reasonable number, GO enrichment analysis is fine. GREAT is a suitable tool for your purpose.