Statistical Methods for Genetic Epidemiology

Columbia Univ New York Morningside, New York NY

Investigators

Linked publications & trials

Abstract

The broad long-term objective is to develop and make accessible statistical methods that remain immune to confounding due to population admixture and stratification while efficiently exploiting the available statistical information in incomplete or phase-unknown genetic data. The health relatedness is that the methods will permit efficient discrimination between genuine evidence for genetic influences on disease and spurious artifacts of miss-specification of population genotype or haplotype frequencies or of population admixture or stratification. A fundamental result that underlies the proposed methods is a representation of efficient unbiased test statistics. The representation is expressed is expressed in terms of a linear regression computation. Methods development will include: deriving the linear regression computation for a variety of hypothesis testing goals; and extending the theory from hypothesis testing to estimation. The hypothesis testing goals will include: testing candidate mutations; testing for mode of inheritance of candidates; discriminating between the effects of linked candidates; using association based methods to detect linkage and using association methods for fine mapping. Two complementary approaches to computational algorithms will be developed: one that involves symbolic solutions to the regression problem and one that uses numerical methods. A rapid application development tool will be used to create a software package with a user-friendly graphical interface and data management and report generation capabilities. The package will be made available on the web as a Windows instillation file.