Case 2   (left ordinate) and (right ordinate) values over 40 trials with larger data. The number of truly differential genes, . The “signal” is very weak (). (a) Subcase 2.1 results. , small relative to (). The plot shows the number of false positives, , over 40 data sets. The corresponding is shown on the secondary ordinate axis. In spite of the relatively weak signal, reranking results in remarkably better performance than the standard approaches, and it produces the discovery list with no false positives in a majority of the 40 trials. (b) Subcase 2.2 results. and      ().