To compare a set of genes is different from all the genes, which test should be used.

Question

[Zhilong Jia]

Here there are two list of genes. One is a set of genes (about 500 genes), the other is all the genes (about 20k). And each gene has a number meaning the number of genes interacted with it. To compare whether or not they are different concerning the number of interacted genes, which kind of test should be used? Thank you.

A = c(4, 5, 40, 3, 0, 0) #length(A)=500
B = c(1,2,0,0,0,43,65,...) #length(B)=20000

Genes from A are also from B.

Answer 1

[russhh]

All normal approximations are out-of-bounds in this setting. I think your best bet is Kruskal-Wallis.

However, I'd temper your enthusiasm for the very low p-values it might return, as there are some serious acquisition biases in interactome data (and KW provides no way to control for them)

A <- rpois(50, 10)
B <- rpois(50, 100)
kruskal.test(list(A, B))

[Zhilong Jia]

Yes, there are biases in interactome data, esp. lots of 0s. Any other test which could be used? Thank you.

[russhh]

Did it not work?

[Deepak Tanwar]

It should work.

[Zhilong Jia]

Yes, it works. Thank you.