Figure 2.

FDR and BUM results from prostate data. (Top) Distribution of the effective false discovery rate (FDR) at the median p-value cutoff when using ten (left) or twenty (right) samples per group in the prostate tumor vs. normal data set. (Bottom) Range of beta-uniform mixture (BUM) models of the distribution of p-values by randomly sampling ten (left) or twenty (right) samples per group. The red (resp., blue) curve marks the random samples with the smallest (resp., largest) estimated number of DEGs.

Zhang and Coombes BMC Bioinformatics 2012 13(Suppl 13):S1   doi:10.1186/1471-2105-13-S13-S1