Electronic Edition @ Stanford.edu (open Access).pdf

The increasing use of Whole Genome Sequencing (WGS) and Whole Exome Sequencing (WES) in clinical settings has made it difficult to determine the causal variant for a given disease, particularly for rare diseases and complex diseases without Mendelian inheritance patterns. To address this challenge, researchers are utilizing a diversity of data sources and algorithms, including model organism phenotyping data, interactome data, orthology, phylogenetic inference, and epigenomics. Recent studies have highlighted the potential of combining genotype, phenotype, epigenetic, and exposure data for precision diagnostics, using methods such as semantic inference and probabilistic modeling. However, these approaches depend on highly curated data and a diversity of software tools and algorithms, and the quality of phenotyping data varies widely. This session includes an invited talk and several papers that explore problems in combining genotype and phenotype data to support rare disease and precision diagnostics, leveraging multiple data modalities such as clinical measures, imaging, natural language processing, and semantics. The papers present various methodologies, including generalized low rank models for capturing phenotypic information within electronic health records (EHRs), diagnosis-guided methods for identifying neuroimaging biomarkers associated with genetic risk factors, and differential network approaches to analyzing metabolomics data.