School Seminars and Colloquia

Prediction of genetic risk of complex disease in individuals using dense, genome-wide genetic markers

Statistics Seminar

by Mike Goddard

Institution: Department of Primary Industries, Victoria
Date: Tue 20th May 2008
Time: 1:15 PM
Location: Room 213 Richard Berry Bldg, The University of Melbourne

Abstract: Most common diseases and most quantitative traits are controlled
by many genes and by environmental factors. The genes affecting
these traits have been difficult to identify. Recently panels of single
nucleotide polmorphisms (SNPs) that cover the whole genome have
become available and have been used in genome wide association
studies that have identified many SNPs that are associated with
risk of disease or other phenotypes. One use of this data is to
predict the future risk of disease of individuals based on the SNP
alleles that they carry. An equivalent use in livestock is to predict
the genetic value of animals from the SNPs so that the best animals
can be selected for breeding. However, typically 100,000's of SNPs
are genotyped on 100's or 1000's of individual people or animals.
The probelm is to estimate a prediction equation based on these
SNPs that will predict the genetic value or phenotype in future
samples of people or animals. Ad hoc model selection methods
work reasonably but the ideal method is to calculate the conditional
expectation for each individual given their SNP genotypes. This
requires a prior distribution of SNP effects on the trait. I will
illustrate the use of different priors with simulated and real data.

For More Information: Guoqi Qian