UW Madison UW School of Medicine and Public Health
UW Biiostatistics & Medical Informatics UW Biostatistics & Medical Informatics UW Madison UW Biostatistics & Medical Informatics Site Map

General Departmental Seminar Series

Gustavo De Los Campos
Agricultural & Applied Economics Animal Science, CALS, University of Madison, Wisconsin

Friday, October 31st, 2008
12:00 pm - 1:00 pm
5275 MSC

Predicting Complex Traits in the Genomic Era

 
ABSTRACT
Nowadays dense marker (e.g. SNPs) panels are available for humans and many plant and animal species. An important challenge is how this information can be brought into models to arrive to prediction of phenotypic outcomes (e.g., growth or disease susceptibility). In the pre-genomic era, a successful approach was to 'regress' phenotypic records on pedigrees using (co)variance structures computed from a pedigree. After a brief review of this approach we look at how marker information can be combined with pedigrees to arrive at predictions of phenotypic outcomes for complex traits.

We evaluate two classes of models: (a) Bayesian Regression coupled with LASSO (BRL), in which regression on marker covariates is treated as in the Bayesian Lasso of Park and Casella (JASA 103: 681-686, 2008) and the remaining components of the model (i.e., regression on pedigree and on other environmental factors) are treated as in the standard quantitative genetic model; and (b) Reproducing Kernel Hilbert Spaces Regression (RKHS) where, as in Gianola and van Kaam (Genetics 163: 347-365, 2008), genomic values are brought into the model using a (co)variance structure computed from the pedigree and marker data simultaneously.

Further we discuss a class of flexible kernels which allow the algorithm to simultaneously pick a kernel and a pattern. Two examples with data involving a self-pollinated collection of wheat lines and an out-cross population of mice will be presented to illustrate the methodologies.

Return to seminar list