Generation of VarData set for the training and testing of MISTIC.

Different filters were applied in order to generate balanced positive set (high-confidence deleterious variants) and negative set (benign variants). a–Selection of variants with a "Pathogenic" information in the clinical significance (CLNSIG) INFO tag in ClinVar VCF file. b–Selection of at least two-stars high-confidence variants with either 'criteria_provided', '_multiple_submitters', 'reviewed_by_expert_panel', 'practice_guideline' or 'no_conflicts' information in the clinical review status (CLNREVSTAT) INFO tag in ClinVar VCF file. c–Selection of high-confidence missense variants with a Disease-Mutation (DM) STATUS INFO tag in HGMD Pro VCF file. d–Selection of missense variants with a depth coverage > 30X and absent from ClinVar and HGMD Pro databases. e–Filtering of variants that overlap any of the training set variants of SIFT, PolyPhen-2, Condel, VEST4, CADD, MetaLR/MetaSVM. (XLSX)