DISCRIMINATOR: Assigning Cohort-Wide Provisional Pathogenicity Classifications to CNVs

Abstract The interpretation of clinical chromosomal microarrays (CMAs) has historically relied on the relevance of identified copy number variants (CNVs) to the clinical phenotype. New interpretation guidelines are focused on standardizing pathogenicity classifications based on genomic location, gene content, and previous publications, rather than the immediate clinical relevance. Here we report on DISCRIMINATOR, which was developed to assign provisional pathogenicity classifications based on genomic location by integrating information on putative benign and pathogenic loci in the human genome. However, its application extends beyond that of a simple classifier. The novel utility of DISCRIMINATOR is its ability to operate on a cohort-level and easily integrate updated definitions of benign and pathogenic regions of the human genome. We used DISCRIMINATOR to assign provisional pathogenicity classifications (‘Benign’, ‘Secondary’, ‘Primary’ or ‘Non-Coding’) to 87,808 CNVs in 3,362 cases ascertained through clinical CMA testing. The majority of identified CNVs were provisionally classified as ‘Benign’ or ‘Non-Coding’ and consistent with their prevalence rates, 15q11.2, 16p11.2, and 22q11.2 were the most common ‘Primary’ CNVs detected. Targeted re-analysis led to the identification of several cases where DISCRIMINATOR identified a ‘Primary’ CNV within a case that had a non-Abnormal CMA test result, and several cases where only benign and/or non-coding CNVs were identified in reports with a ‘VUS’ CMA test result. Together these results show the utility of large-scale re-analysis of CMA data and how DISCRIMINATOR addresses this long-standing challenge.