A Novel Approach with an Optimal Power for Joint Multiple-QTL Mapping of Complex Traits

• Activity Number: 139
• Type: Contributed
• Date/Time: Monday, August 4, 2014 : 8:30 AM to 10:20 AM
• Sponsor: Section on Statistics in Epidemiology
• Author(s): Riyan Cheng*+ and Justin Borevitz
• Companies: and Australian National University
• Keywords: model selection ; multitrait mapping ; quantitative trait locus ; seemingly unrelated regression equations ; statistical power

Abstract

Joint analysis of multiple correlated traits (multitrait) has been advocated for quantitative trait locus (QTL) mapping; however, its usefulness is not fully exploited. We have developed powerful statistical procedures and computational strategies for genome-wide association studies under the framework of multitrait multiple-QTL models which, unlike existing models, can associate different sets of QTL with different traits and thus facilitate biological interpretation of results. We investigate various methods to exclude non-significant QTL effects, including backward elimination with dynamic parameter penalties. Our proposed multitrait methodology achieves an optimal statistical power for QTL identification and thus potentially improves the chance of genetic discoveries, and is widely applicable to genetics studies of human diseases and animal/crop breeding. We develop a computationally efficient algorithm, which takes advantage of both matrix sparsity and dimension reduction, for model parameter estimation so it is suitable for our multitrait multiple-QTL mapping where model selection is involved and the computation is intensive.