Table 2.

Statistical comparisons of prediction performance.

Method	Method	Set	n	p-value
Jones-UCL	Priors	MF	366	7.75E-06
Jones-UCL	BLAST	MF	351	3.37E-23
Jones-UCL	Priors	BP	436	3.43E-15
Jones-UCL	BLAST	BP	387	6.62E-12

The performance of pairs of methods was compared separately on the two benchmark subsets MF and BP. We tested the null hypothesis that the distributions of the COGIC scores over the set of common targets were indistinguishable with a paired two-sided Wilcoxon sum rank test. The table reports the number n of common targets and the p-value.