Table 1. Evaluation of feature subsets (ROC-AUC, 10-fold cross-validation with linear support vector machine classifier) to discriminate between CPDs versus SNPs and disease-causing mutations versus common SNPs (DM versus SNPs).

Feature subset	CPD-AUC (%)	DM-AUC (%)	Performance reduction for CPD (CPD-AUC versus DM-AUC)
Genomic MSA	74.0	94.6	−20.6
Protein MSA (homologous)	60.4	80.7	−20.2
Local protein structure	56.5	68.5	−12.0
Regional protein composition	55.3	63.5	−8.2
Exonic features	64.7	71.0	−6.4
Annotated functional sites	50.3	55.3	−5.0
Amino-acid features	64.5	69.1	−4.6
Random value (control)	50.0	48.7	1.3

Features ranked by performance reduction between CPD set and DM set.