ROC curves for two non-coding RNA predictors, ClosingBp (Bradley RK, Uzilov AV, Skinner M, Bendaña YR, Barquist L and Holmes I, submitted) and EVOFOLD  (implemented using XRATE), comparing DAWG  to the richest GSIMULATOR and SIMGENOME models. The three curves for each gene predictor clearly illustrate that increased model richness (DAWG → GSIMULATOR → SIMGENOME) yields higher estimated FPR. See the caption to Figure 1 for an explanation of why the asymptotic sensitivity is less than 1.0.
Varadarajan et al. Genome Biology 2008 9:R147 doi:10.1186/gb-2008-9-10-r147