User-Item Recommendation Approaches to Detect Genomic Variant Interactions

The genome of an organism is comprised of the nucleotides A, G, C, and T. Genomic mutations often result in phenotype changes and are primarily classified as structural variants or single nucleotide polymorphisms (SNPs). Epistasis describes when two or more mutations are responsible for changes in qualitative or quantitative phenotypes in an organism. While some SNPs are responsible for diseases, the interaction of three or more mutations and their affect on quantitative phenotypes are not completely understood. Past methods often require Bonferroni or similar approaches to account for the exponential number of groupings of mutations. In this work, we propose user-item recommendation approaches for genomic signals to reduce the search space for higher-order interactions. We present results in the context of muscle mice mutations and are able to identify statistically significant groupings related to their body weight.