Advantage of the pooling approach.

Average probabilities (over 100 trials) that random, withheld gene-disease links support nodes more significant by the traditional method () or the proposed pooling method () for the 9 query gene sets. Significance in each trial was computed without the withheld links. When is larger than , the nodes that are more significant under the pooling approach tend to be more consistently supported by the withheld data, our proxy for correctness.