GO term size influences the performance of individual methods. GO term size appears to be a significant factor that influences the relative performance among groups (the performance of each group is shown by the average AUC values). Although there is no general trend common to all groups, individual groups demonstrate consistent trends regardless of what GO branch was tested. For example, groups B and D perform better on big GO terms. Group C performs better on small terms, and groups E to H perform well on intermediate-sized GO terms. Asterisks indicate second round submissions. AUC, area under receiver operating characteristic curve; BP, biological process; CC, cellular components; GO, Gene Ontology; MF, molecular functions.