Motivation      

Genotyping and large-scale molecular phenotyping are already available for large patient cohorts and will soon become routinely available for all patients. These data are setting the stage for rapid advances in personalized medicine, enabling better disease classification, more precise treatment, and improved screening for disease prevention.

Robust statistical and computational methods for analyzing these data are critical to realizing the promises of personalized medicine. Important analysis problems include: identifying and correcting for hidden structure; dealing with missing data, data heterogeneity; and addressing the problem of multiple testing. For example, in Genome-Wide Association studies (GWAS), population structure and family relatedness can reduce power and cause spurious associations. In gene expression studies, experimental artifacts or environmental influences, have been shown to corrupt results of naïve analyses. Additionally, hypothesizing and inferring hidden phenotypes, such as cell type or transcription factor activity help reveal hidden structure. Further advances in statistical modeling and machine learning are still needed to realize the promise of personalized medicine of delivering a "computed therapy".

This session focuses on methods to address open and new methodological problems pertaining to genotype data, intermediate phenotypes, clinical variables and disease. Problem areas within the scope include methods for Genome-Wide Association Studies, gene expression modeling with relevance to disease and causal modeling integrating information from genotype, intermediate phenotype and disease labels. The session is intended to have a broad target audience including method developers and practitioners in the fields of medical and human genetics, statistical genetics and related areas.

Session Topic