Fig. 4

Benchmark of GRIPT, VAAST2, CMC, SKAT, and KBAC on 400 Mendelian disease genes. The AR and AD models were tested with 0.5%, 1%, 2%, and 3% of patients carrying the pathogenic mutations of each of 200 AR genes and each of 200 AD genes, respectively. The patient cohort size was 600. The control cohort size was 5000. The performance of GRIPT, VAAST2, CMC, SKAT, and KBAC are shown in red, blue, green, purple, and orange, respectively. a The ranking of 200 AR genes. b The power of the five tests for 200 AR genes. c The number of significant autosomal candidates under the AR model. d The ranking of 200 AD genes. e The power of the five tests for 200 AD genes. f The number of significant autosomal candidates under the AD model. The rankings of AR/AD genes generated by GRIPT were compared to those generated by the other four methods respectively with one-tailed WRST. The methods that generated significantly worse ranking than GRIPT were marked with “*” if p value < 0.05, “**” if p value < 0.01, and “***” if p value < 0.001