Table 2.
Comparison between predictors with validation dataset of 1,100 mutations (23 damaging + 1,077 neutral)
| MITOCLASS.1 | POLYPHEN-2 | PROVEAN | MUTPRED | |
|---|---|---|---|---|
| Sensitivity | 95.7 | 91.3/94.7 | 91.3/87.7 | 60.9/57.9 |
| Specificity | 58.7 | 47.7/46.9 | 60.4/59.2 | 85.6/87.3 |
| TP | 22 | 21/54 | 21/50 | 14/33 |
| TN | 623 | 514/1303 | 650/1646 | 922/2426 |
| FP | 454 | 563/1475 | 427/1132 | 155/352 |
| FN | 1 | 2/3 | 2/7 | 9/24 |
For PolyPhen-2, Provean and Mutpred, the complete mdmv.1 dataset has also been analyzed (numbers after the slash). PolyPhen-2 is unable to predict the phenotype of 10 missense mutations because much of the initial and final sequence of p.MT-ND5 is non-aligneable due to large stretches of repeats and/or high compositional biases as commented by authors. For the sake of comparison, we consider these unknown predictions as neutral variants. TP, TN, FP, FN refers to true positives, true negatives, false positives and false negatives respectively. Sensitivity is estimated as [TP/(TP + FN)], specificity as [TN/(TN + FP)]