Automating ACMG Variant Classifications Using BIAS-2015: An Algorithm Overview and Benchmark Against the FDA-Approved eRepo Dataset

In 2015, the American College of Medical Genetics and Genomics (ACMG) in collaboration with the Association of Molecular Pathologists (AMP) published guidelines for the interpretation and classification of germline genomic variants. The ACMG terminology guidelines outlined criteria for assigning one of five categories: benign, likely benign, uncertain significance, likely pathogenic and pathogenic. While the paper laid out 28 different classifiers and the justification for them, it did not provide specific algorithms for implementing these classifiers in an automated manner. Here we present the Bitscopic Interpreting ACMG Standards 2015 (BIAS-2015) software as a complete, open-source algorithm which categorizes variants according to the ACMG classification system. BIAS-2015 evaluates 18 of the 28 ACMG criteria to classify variants in an automated and consistent way while recording the rationale for each classifier to enable in-depth review. We used the genomic data from the ClinGen Evidence Repository (eRepo v1.0.29), one of two FDA-recognized human genetic variant databases, to evaluate the performance of the BIAS-2015 algorithm. All code for BIAS-2015 has been made available on GitHub.