Gene Set Enrichment by Fisher's Exact Test

Gene set enrichment by Fisher's Exact Test evaluates if the proportion of interesting genes in a pathway (or gene set) is enriched relative to the proportion of interesting genes not in a pathway.

• It uses a 'competitive null hypothesis' rather than 'self-contained'.
• It can lose control over alpha when genes are not independent.
• It can also have a bias based on pathway size.
• It remains the most commonly used gene set enrichment test.

Based on a 2x2 table of genes counts split by whether the genes are of interest or not, and whether the genes are in a pathway or not, a p-value from Fisher's exact test can be calculated:

a <- 17; b <- 181; c <- 449; d <- 19551 
fisher.test(matrix(c(a,b,c,d),2,2,byrow=TRUE), 
            alternative='greater')$p.value

[1] 3.693347e-06

The Shiny app implements this code in an easy to use interface.

Improvements to consider:

• User should be able to paste lists of genes
• User should be provide measures of effect size (e.g., odds ratio)
• User should be offered tests which allow gene dependence