485
Approximation of Reliability of Direct Genomic Breeding Values

Monday, August 18, 2014
Posters (The Westin Bayshore)
Mehdi Sargolzaei , The Semex Alliance, Guelph, ON, Canada
Larry R Schaeffer , Centre for Genetic Improvement of Livestock, University of Guelph, Guelph, ON, Canada
Jacques P Chesnais , The Semex Alliance, Guelph, ON, Canada
Gerrit Kistemaker , Canadain Dairy Network, Guelph, ON, Canada
George R. Wiggans , Animal Improvement Programs Laboratory, Agricultural Research Service, USDA, Beltsville, MD
Flavio S Schenkel , Centre for Genetic Improvement of Livestock, University of Guelph, Guelph, ON, Canada
Abstract Text:

Two methods to efficiently approximate theoretical genomic reliabilities are presented. The first relies on the direct inverse of the left hand side (LHS) of mixed model equations. It uses the genomic relationship matrix for a small subset of individuals with the highest genomic relationship with the individual of interest. The second is a ridge-regression method using the direct inverse of LHS for a small subset of SNP. The performance of the methods was evaluated for the North American genomic data set, consisting of 228,168 genotyped individuals. The ridge-regression method gives very high correlations between theoretical and estimated reliabilities for subsets of 5k SNP and above. It is easily applicable to large data sets. Both methods lead to some biases in the mean and SD of reliabilities but these can be corrected by pegging to theoretical values or values from validation studies.

Keywords:

direct inverse

genomic selection

reliability approximation