Table 5 Performance on 72 GCK variants with clear clinical annotations taken from ClinVar [43], using all predictors assessed as well as experimental fitness data obtained by yeast complementation assays [52]. The best score for each metric is indicated in bold
From: FiTMuSiC: leveraging structural and (co)evolutionary data for protein fitness prediction
Method | Sensitivity | Specificity | BACC | AUC-ROC |
|---|---|---|---|---|
Experimental | 0.59 | 1.00 | 0.80 | 0.732 |
FiTMuSiC | 0.78 | 1.00 | 0.89 | 0.990 |
FATHMM | 1.00 | 0.00 | 0.50 | 0.766 |
PROVEAN | 0.84 | 1.00 | 0.92 | 0.976 |
DEOGEN2 | 0.97 | 1.00 | 0.99 | 0.995 |
PolyPhen-2 | 0.93 | 1.00 | 0.96 | 0.978 |
EVE | 0.58 | 1.00 | 0.79 | 0.807 |
Sequence UNET | 0.52 | 0.67 | 0.59 | 0.638 |
MutPred2 | 0.91 | 0.67 | 0.79 | 0.865 |
AlphaMissense | 0.87 | 1.00 | 0.93 | 0.995 |