Integration of Omic Data in the Analysis of Gene x Environment Interaction

Project 2: Integration of Omic Data in the Analysis of Gene x Environment Interaction Abstract The availability of high-volume ‘omic’ data, including gene expression, metabolome, methylation, and microbiome, provides exciting opportunities to identify novel gene-environment (G×E) and omic × E interactions affecting cancer and other complex traits. For example, the FIGI colorectal cancer consortium has generated transcriptomic (gene expression) data on both normal tissue and colon organoids to inform the discovery of G×E and expression × E interactions for colorectal cancer in a sample of over 130,000 cases and controls, with exposure data on established risk factors including tobacco, alcohol, obesity, and red meat. The multi-ethnic cohort includes over 215,000 subjects followed for up to 30 years, with biomarkers, metabolomic, and microbiome data available on selected subsamples and nested case-control samples of breast and colorectal cancer. In addition to potentially improving power for identifying novel interactions, the use of omic data holds promise to inform the biological mechanisms by which genes and exposures affect a particular trait. This project will develop two types of novel methods that leverage omic data to identify interactions in a genomewide scan. The first (Aim 1) considers one factor at a time (e.g. one SNP, one gene) and uses novel two-step screening/testing methods to discover G×E or omic × E interactions. The second (Aim 2) approach is a joint model considering SNPs and omic data simultaneously, using novel hierarchical modeling techniques to guide the discovery of G×E and omic × E interactions. For both Aims 1 and 2, we will consider the various types of exposure data that may be available, ranging from simple yes/no indicators from questionnaires to integrated exposure measures constructed using statistical models, with or without relevant omic data. Aim 3 will focus on applying the methods from Aims 1 and 2 to several cancer-related data resources, including epidemiological investigations such as FIGI and MEC and a clinical trial examining modifiers of treatment outcomes in colorectal cancer patients. Overall, this project will develop statistical methods to use both integrative omic and environmental exposure approaches to improve power for identifying novel G×E and omic × E interactions as well as to inform the biological mechanism by which these factors affect the risk or prognosis of cancer. We will leverage our collaborations on several cancer-related studies to guide our methods development process, to design realistic simulation studies for evaluating the methods, and to assure that methods we develop are translated into real-data applications.