Statistical Methods for Cancer Genes

点击翻译按钮获取中文摘要

DESCRIPTION (provided by applicant): Identifying individuals at high risk of cancer because of inherited genetic susceptibility is complex and increasingly important. Probabilistic prediction algorithms that exploit domain knowledge of Mendelian inheritance and other biological characteristics of susceptibility genes successfully contribute to improved screening, prevention, and genetic testing, and to the design and analysis of cancer studies. The investigators have developed, validated, applied and disseminated the widely used Mendelian model BRCAPRO. Based on their experience they have identified the need for a new generation of Mendelian prediction models in cancer genetics. The first aim will develop statistical approaches that generalize Mendelian models currently used in clinical genetic counseling practice. Innovation will focus on five areas: A) accounting for errors in reported pedigrees; B) accounting for dependencies in time-to-event distributions for multiple cancer sites; C) accounting for familial correlations arising from shared environmental factors or other sources; D) accounting for multiallelic syndromes; and E) incorporating information on covariates and on biomarkers related to the genes' activity. The second aim will introduce a novel class of multi-syndrome models to simultaneously identify cancer syndromes and predict mutation carrier status. These will enable clinicians and researchers to address the emerging challenges posed by the overlap in phenotype for cancer susceptibility genes, and by the high frequency of sporadic families with multiple sites. The third aim will develop flexible user-friendly software for the application of the methods in both research and clinical settings. The aims of this proposal will overcome pressing practical limitations of tools currently used in clinical genetic counseling, and thus contribute to improved screening, prevention and decision making about genetic testing.