Tuesday, November 3, 2009: 11:15 AM
Convention Center, Room 405, Fourth Floor
Peter Bradbury, USDA-ARS, Ithaca, NY, Edward S. Buckler, USDA-ARS, Cornell Univ., Ithaca, NY and Zhiwu Zhang, Institute for Genomic Diversity, Cornell Univ., Ithaca, NY
Abstract:
Mixed linear models have been shown to be effective in reducing false positives and in increasing power when applied to association mapping. These models accomplish that by compensating for population stratification and relatedness between taxa. Within the mixed model framework, population structure can be represented as either a fixed effect, a random effect or both. The resulting analysis generates estimates of genetic variance, error variance, and heritability and provides tests of significance for genetic markers. Care must be taken both in how the population structure is represented in the model and in interpretation of the results. The nature of these issues and how they are handled by TASSEL software is discussed.