CAREER:Resources for mixed model association mapping of complex traits

Harvard University, Cambridge MA

Investigators

Abstract

The goal of this project is to devise new statistical methods for identifying genetic variants associated to disease resistance and other traits in plants, livestock, and model organisms such as fruit flies and mice. Specifically, the research will focus on mixed model association methods, a class of methods in which association statistics are computed by considering all genetic variants simultaneously in order to increase power to detect true associations and to avoid false-positive associations. Although mixed model association methods have previously been applied to studies in these species, existing methods are based on statistical assumptions that are unlikely to hold in real data sets, particularly in the large whole-genome sequencing data sets of the future. The new methods produced by this research will address those limitations. The investigator will use the research activities in this project to help teach and train numerous students, including those from underrepresented groups. The project will release publicly available, open-source software implementing the new methods. Making these statistical methods and software widely available to other researchers is likely to lead to new scientific discoveries and insights when the methods are applied to large whole-genome sequencing data sets. The research will produce new mixed model association methods and software for application to complex trait association studies in plants, livestock and model organisms that will address limitations of existing methods, increasing power while controlling false-positive associations. Specifically, the methods developed will account for general genetic architectures, moving beyond the "infinitesimal" genetic architecture with all markers associated that underlies existing mixed model association methods; will account for non-randomly ascertained traits; and will enable pooled analyses of rare variants in a mixed model framework. All of these new features are expected to increase statistical power to detect true associations. The broader impacts of this project will include: enhanced computational infrastructure for conducting mixed model association mapping via the continued release of open-source software; multi-disciplinary synergy from research, collaborations and conferences that will bring together plant, livestock, model organism and human geneticists; and increased participation of underrepresented groups via the HSPH Summer Program in Quantitative Sciences. Software produced by this project will be available at http://www.hsph.harvard.edu/alkes-price/software/.