How is the “Empirical non-random distribution p-value” calculated?
A clusteredness metric is defined as the mean (over genes) of the city-block map distance from a gene to the closest other gene (small values are better). The clusteredness metric is calculated for the genes in question (e.g. with a given GO term). Then the same metric is calculated for 100 random gene sets of the same size. The p-value is the fraction of random samples that are more clustered than the real sample. If no random samples are better clustered then the p-value is not zero, but must be less than 0.01.
