Statistical Methods for Integrative Genomics in Cancer

点击翻译按钮获取中文摘要

DESCRIPTION (provided by applicant): We aim to develop novel statistical methods to address some of the major problems facing cancer genetic epidemiologists in the "post-GWAS" era and to illustrate their use for discovery of novel biology in various colorectal cancer (CRC) studies. These methods leverage prior biological knowledge to inform integrative genomics analyses (Project 1), use phylogenetic information to infer gene function as inputs to our epidemiologic modeling projects (Project 2), model the role of the microbiome and the exposome in cancer risk (Project 3), and exploit intra-tumor heterogeneity to learn about somatic tumor evolution and how this process is modified by the internal environment (Project 4). These four projects will be supported by an administrative core and three shared resource cores on functional annotation, high performance computing, and software development. The entire program is motivated by an overall objective of providing tools for evaluating the impact of potential preventive or therapeutic interventions based on modifiable risk factors. Specifically, the aims of the overall program are (1) to develop statistical analysis methods to integrate multiple types of omics data that describe both constitutional and acquired genomic variation as well as measures of the external and internal environment into comprehensive risk prediction models, leveraging external information; (2) to apply these methods to various studies of CRC etiology and prognosis to uncover novel associations and to develop predictive models that would have translational significance for possible primary, secondary, and tertiary interventions; and (3) to establish an infrastructure (administrative, bioinformatic, computational, software) to support the various research projects and facilitate making our methods accessible to the broader scientific community. This will be achieved by a combination of theoretical developments, simulation studies closely keyed to real data projects, applications to several studies of CRC, and distribution of software for use by outside investigators. Beyond applications to colorectal cancer, our methods will be broadly applicable to other cancer types and many other chronic diseases.