Validation results for five features sets.

Weighted F scores are reported on the validation set (20% HumVD-train) for all algorithms. ‘Sequence’ refer to sequence-based features (PSIC, Nobs, MinDJxn, CodPos and SeqCons); ‘Structure’ refers to structure-based features (KDmean, RSAmax, nRSA, nNum, Bstddev, nB, nKD, nSC, nBinding, Binding, Mapreg); ‘Str no Neigh’ refers to structure-based features without neighbour information (KDmean, RSAmax, Bstddev, Binding, Mapreg); ‘Seq + Str’ refers to all sixteen sequence and structure features listed above; ‘All + mutation’ refers to all prior features listed above and features based on difference between wild-type listed in S1 Table. The boldface numbers highlights top two weighed F scores for each feature sets.