Principal Investigator: Professor Nicholas Schork
The Translational Genomics Research Institute (Arizona, USA)Tags: 43026, association tests, causal models, clustering, Heterogeneity, mediation, mixture distributions
No two individuals are exactly alike. This is particularly true with respect individuals’ genetic, physiologic, exposure and behavioral profiles. These differences create variation in individual disease susceptibility as well response to interventions. This suggests that two individuals may have the same disease but for different reasons, or respond differently to treatments for that disease. Such heterogeneity makes it difficult to identify factors that contribute to disease and treatment response that might be used to assess an individual’s disease risk or determine what treatments they might benefit from. In order to overcome this, researchers must be able to: 1. Identify evidence for such heterogeneity in relevant data sets harboring individual genetic and phenotypic information; and 2. Account for it in specific analyses relating genes and phenotypes in those data sets. Although some methodologies exist for identifying and characterizing heterogeneity in biomedical data sets (e.g., cluster analysis methods), they have not been used widely in the specific contexts of which I am interested in, including genetic association analyses, tests of causal pathways leading from genes to phenotypes, and cross-phenotype analyses. Identifying and accommodating heterogeneity in such analysis contexts may reveal novel insights into how diseases develop and how they can be treated and managed.