GO::TermFinder will take a list of genes (any name that appears in an annotation file will do, though there are caveats for ambiguous names), an ontology file, and an annotation file, and determine the P_values for the observed frequency of annotation to those genes for each GO node. There are currently some differences between how the P-values are
The full announcement can be read here: http://bio.perl.org/pipermail/bioperl-l/2003-March/011707.html