ClinVar and HGMD data analysis

Disease-associated variants were selected from the NCBI ClinVar database (downloaded September 9, 2017)³⁰ and the Human Gene Mutation Database (publicly available variant data before 2014.3)³¹ for computational screening using the inDelphi models trained on U2OS, HCT-116, and HEK293 cells¹². ClinVar variants were included where at least one submitting lab designated the clinical significance as ‘pathogenic’ or ‘likely pathogenic’ and no submitting lab had designated the variant as ‘benign’ or ‘likely benign’. HGMD variants were included with any disease association with the HGMD classification of ‘DM’ or disease-causing mutation. SpCas9 gRNAs and their cleavage sites were enumerated for each disease allele. Genotype frequency was predicted for each tuple of disease variant and unique cleavage site (Supplementary Data Files ⁷ and ⁸). For each unique variant, the single best gRNA was identified as the gRNA inducing the highest predicted frequency of repair to wild-type genotype (for 1-bp deletion variants) or allele installation (for 1-bp insertion variants).