Bayesian Modeling and Optimal Design for Studies of Gene-Environment Association

点击翻译按钮获取中文摘要

DESCRIPTION (provided by applicant): Success of modern epidemiological investigations into complex disease and, in particular, success of the Genes and Environment Initiative (GEI) will depend on development of methods that address these challenges and on their efficient, open source implementation. Our objective with the proposed body of research is to utilize Bayesian statistical approaches to identify optimal designs and develop efficient analytic strategies for studies of association between genes, environment and disease. This research will be organized into three aims. The focus of Aim 1 will be to develop algorithms and methods to determine Bayesian optimal study designs for gene by environment studies whose purpose is either discovery or replication of associations. This will include algorithms for determining optimal multi-phase study designs for discovery of associations given a set of resource constraints and optimal designs and methods of analysis for consortium-based validation studies. The focus of Aim 2 will be to develop methodological approaches to the analysis of data generated by both hypothesis driven and unbiased (i.e. genome-wide) association studies. This work will include methods for detection of candidate pathway and candidate pathway-environment interaction associations accounting for data on pathway structure and function, if it exists and computer efficient methods for genome-wide analysis of gene-environment interaction. Aim 3, Software Development, will focus on development of efficient, portable software implementations of the methods developed in context of Aims 1 and 2. At the conclusion of the proposed work, we will have developed, coded and tested algorithms or approaches to study design and analysis of data generated in studies of genes, environmental exposure and disease. Finally, to ensure the accessibility of our work, we will package the software we develop and its documentation in a user-friendly R statistical language package and maintain it on the Comprehensive R Archive Network (CRAN). (End of Abstract)