Fig. 2: Phrank performance on a set of 169 diagnosed patients.
a The fraction of cases correctly diagnosed (y-axis) is plotted against a cumulative causal disease rank (x-axis: is the correct disease diagnosis ranked first? Is it ranked in the top 5? etc.) for both Phenomizer and Phrank. Phrank outperfoms Phenomizer significantly (p ≤ 2.2*10−16, Wilcoxon signed rank test) as Phrank ranks the causative disease higher than Phenomizer at all ranking thresholds. b The fraction of cases correctly diagnosed (y-axis) is plotted against a cumulative causal gene rank (x-axis: is the correct gene diagnosis ranked first? Is it ranked in the top 5? etc.) for Phevor, PhenIX, and Phrank. Phrank slightly outperforms PhenIX and Phevor (p ≤ 0.19 and p ≤ 0.41, Wilcoxon signed rank test, respectively), but the difference is not statistically significant. Over all cases Phrank gives the causative gene an average rank of 9.5, while PhenIX and Phevor achieve average ranks of 15.5 and 21, respectively
