Artificial Intelligence and Geographic Analysis of Clinical Genetic Data in California’s Central Valley

Abstract

Abstract Introduction Data science methodologies can be utilized to ascertain and analyze clinical genetic data that is often unstructured and rarely used outside of patient encounters. Methods Genetic variants from all genetic testing resulting to a large pediatric healthcare system for a five-year period were obtained and variants were reinterpreted utilizing the Franklin© Artificial Intelligence (AI). Utilizing PowerBI©, the data were further matched to patients in the electronic healthcare record and matched to demographic data to generate a variant data table and map variants as a choropleth. Results Three thousand sixty-five variants were identified and 98% were matched to patients with geographic data. Franklin© changed the interpretation for 27% of variants. A total of 723 Mendelian genetic disorders were identified with disorder prevalence estimation. Mapping of variants demonstrated hot-spots for pathogenic genetic variation such as PEX6-associated Zellweger Spectrum Disorder. Seven patients were identified with Bardet-Biedl syndrome and seven patients with Rett syndrome amenable to newly FDA-approved therapeutics. Discussion Utilizing readily available software we developed a database and Exploratory Data Analysis methodology enabling us to systematically reinterpret variants, estimate variant prevalence, identify patients amenable to new treatments, and localize geographies enriched for pathogenic variants.