two populations. Using this cutoff, we computed a one-sided Fisher's exact probability to determine enrichment of GO terms in the top 5% of probes. For a single population, such as CEU, all GO term p-values from pairwise population comparisons with CEU were combined using the Fisher combined probability and compared to the combination of GO term p-values for the other populations (in this case, excluding CEU). We then filtered from the primary population those GO enrichment terms which were also found significantly differentiated among other populations. From this filtered list, we then selected the top ten GO terms for each population. These top ten values were used as indicative of functions which are significantly differentiated between the primary population and other populations and not between any other populations.
