Figure 3

Performance statistics of the top five mutation effect prediction algorithm combinations as ranked by composite scores. Prediction results of the non-neutral (n = 849) and neutral (n = 140) single nucleotide variants (SNVs) in the entire dataset (A, B) and the non-neutral (n = 188) and neutral (n = 109) SNVs not present in the COSMIC dataset (C, D) are shown. Results are ranked according to the composite scores of each mutation effect prediction algorithm combination, and the corresponding accuracy, sensitivity, specificity, positive predictive value (PPV), negative predictive value (NPV), and composite score of the top five prediction algorithm combinations in subset 1 (A, C) and subset 2 (B, D) are plotted. Error bars represent the 95% CIs generated by 1,000 random samples of subsets 1 and 2. Red bars represent predictor combinations, blue bars single/independent predictors, and orange bars meta-predictors. Blue stars: statistically significant improvement in composite score as compared to that of the best performing single/independent predictor; orange stars: statistically significant improvement in composite score as compared to that of the best performing meta-predictor; blue triangles: statistically significant improvement in NPV as compared to that of the best performing single/independent predictor; orange triangles: statistically significant improvement in NPV as compared to that of the best performing meta-predictor.