Skip to main content

Table 3 Performance statistics of mutation effect prediction algorithms using all single nucleotide variants tested functionally (n = 989)

From: Benchmarking mutation effect prediction algorithms using functionally validated cancer-related missense mutations

Prediction algorithm Accuracy (95% CI) Sensitivity (95% CI) Specificity (95% CI) PPV (95% CI) NPV (95% CI) Composite (95% CI)
CHASM (breast) 88.68% (86.55-90.70%) 89.99% (87.81-91.92%) 80.71% (73.33-87.12%) 96.59% (95.20-97.73%) 57.07% (50.23-64.58%) 3.2436 (3.1127-3.3669)
CHASM (lung) 90.09% (88.27-91.81%) 92.23% (90.40-94.04%) 77.14% (69.93-83.58%) 96.07% (94.74-97.31%) 62.07% (54.72-69.38%) 3.2751 (3.1418-3.3988)
CHASM (melanoma) 89.69% (87.77-91.51%) 93.64% (92.00-95.37%) 65.71% (57.35-73.51%) 94.31% (92.67-95.76%) 63.01% (55.40-70.92%) 3.1667 (3.0181-3.3079)
FATHMM (cancer) 91.00% (89.18-92.62%) 97.88% (96.87-98.82%) 49.29% (41.08-57.24%) 92.13% (90.26-93.72%) 79.31% (70.59-87.66%) 3.1860 (3.0326-3.3292)
FATHMM (missense) 82.51% (79.98-84.73%) 87.75% (85.46-89.94%) 50.71% (42.98-58.27%) 91.52% (89.53-93.23%) 40.57% (33.53-47.72%) 2.7056 (2.5576-2.8521)
Mutation Assessor 74.70% (71.89-77.51%) 76.65% (73.87-79.65%) 62.86% (54.61-71.21%) 92.59% (90.35-94.55%) 30.77% (25.69-36.70%) 2.6287 (2.4835-2.7842)
MutationTaster 88.57% (86.65-90.60%) 90.81% (88.94-92.84%) 75.00% (66.91-82.00%) 95.66% (94.10-97.00%) 57.38% (50.00-65.14%) 3.1885 (3.0502-3.3300)
PolyPhen-2 79.78% (77.05-82.11%) 84.22% (81.83-86.61%) 52.86% (44.51-61.15%) 91.55% (89.45-93.40%) 35.58% (29.44-42.16%) 2.6420 (2.4895-2.7915)
PROVEAN 73.71% (70.88-76.64%) 74.44% (71.51-77.43%) 69.29% (61.72-76.06%) 93.63% (91.82-95.31%) 30.89% (25.93-36.56%) 2.6825 (2.5505-2.8088)
SIFT 80.99% (78.26-83.42%) 86.10% (83.89-88.45%) 50.00% (41.93-58.40%) 91.26% (89.08-93.15%) 37.23% (30.34-44.28%) 2.6460 (2.4895-2.7968)
VEST 75.03% (72.50-77.76%) 82.69% (80.28-85.24%) 28.57% (21.01-35.40%) 87.53% (85.12-89.75%) 21.39% (15.91-27.57%) 2.2018 (2.0568-2.3395)
CanDrA (breast) 81.81% (79.23-84.26%) 95.38% (93.85-96.81%) 0% (0-0%) 85.19% (82.83-87.30%) 0% (0-0%) 1.8056 (1.7765-1.8340)
CanDrA (lung) 91.28% (89.54-93.04%) 92.86% (91.05-94.62%) 80.95% (73.64-87.32%) 96.97% (95.75-98.03%) 63.35% (56.10-71.01%) 3.3413 (3.2103-3.4747)
CanDrA (melanoma) 88.97% (87.02-90.91%) 90.73% (88.69-92.81%) 77.60% (69.57-84.40%) 96.33% (94.89-97.53%) 56.40% (48.77-63.74%) 3.2105 (3.0705-3.3507)
Condel 85.84% (83.42-88.17%) 92.58% (90.81-94.28%) 45.00% (37.06-53.72%) 91.08% (89.16-92.94%) 50.00% (41.43-59.06%) 2.7866 (2.6222-2.9552)
  1. Based on the prediction results of all non-neutral and neutral single nucleotide variants included in this study, the accuracy, sensitivity, specificity, positive predictive value (PPV), negative predictive value (NPV), and composite score for each predictor were computed. The 95% confidence intervals (CI) generated by bootstrapping are shown in parentheses.