Testing overlap significance of three gene sets using the hypergeometric distribution

Question

[Pac314]

Is it appropriate to take a three-way intersection of three overlapping gene sets and use the phyper function in R to assess the significance of the overlap? I have used the hypergeometric distribution for pairwise intersections before but not for > 2 intersections.

[rpolicastro]

You're better off just simulating it. For 1000+ iterations randomly sample 3 gene sets of equal size to your original 3 gene sets and record the overlap length. Your observed overlap should be greater than 95% of these simulated overlaps.

[Papyrus]

+1 for the simulation mentioned by rpolicastro , and it its the typical approach. However also check out the SuperExactTest R package, which implements testing for intersections of multiple sets (it is also nice to analize all possible intersections in one go).

[Pac314]

Thank you both for your answers! This library is perfect for my analysis. Thank you for sharing this!