Prediction performance of feature subsets on test data.

aAll scores are calculated for the variations that were predicted at confidence level 0.95.bSens, Sensitivity; Spec, Specificity; Acc, Accuracy; OPM, Overall performance measure; AA, Amino acid features; GO, GO annotation derived feature; SelPres, Selective pressure; SeqProf, Sequence profile features (proportion of reference amino acid, proportion of variant amino acid and number of sequences in the multiple sequence alignment)cCoverage is the proportion of the data that are predicted either pathogenic or neutral.Prediction performance of feature subsets on test data.