Performance comparison of the original CoSBI and semi-supervised CoSBI (SS-CoSBI) algorithms across a range ofalphaandbetaparameter settings. X-axis, bicluster sets of different sizes generated by the two algorithms using different parameter settings. Y-axis, the enhancer enrichment scores for each bicluster set. First, a hypergeometric p-value was computed for each bicluster in a set. Then, the enrichment score for each set was computed as the minus logarithm (base 2) of the average hypergeometric p-values of each bicluster set. Values around the SS-CoSBI curve are KS test p-values of enrichment score distributions. Each pair of bicluster sets of the same or similar size generated by CoSBI and SS-CoSBI was compared.
Teng and Tan BMC Genomics 2012 13:301 doi:10.1186/1471-2164-13-301