Medical Big Data Analysis System to Discover Associations between Genetic Variants and Diseases

Electronic Health Records (EHRs) and genomic data are fruitful resources to find associations between genetic variants and human diseases. Medical researchers have used Phenome-Wide Association Study (PheWAS) to find diseases relevant to genomic variants, and Genome-Wide Association Study (GWAS) to discover genomic variants related to specific diseases. In the era of Big data, the quantity of disease and genomic data have significantly increased, resulting in long processing and research time for Big data analysis. We designed and implemented a prototype of Medical Big Data Analysis System that finds the links between genetic variants and human disease with PheWAS and GWAS dataset of a population-based biobank in Marshfield Clinic Institute. The system consists of front-end Web-based graphic user interface using Shiny application, and back-end Big data analysis servers using Apache Spark and Mongo database. We evaluated the associations with selected variants and diseases and present the performance and CPU/memory overhead of the system while varying the number of executors in Spark cluster, which provides readers with multiple factors for deploying medical Big data analysis system.