Genetic evaluation including phenotypic, full pedigree and genomic data

Abstract

Genomic evaluations could be obtained using a unified methodology that combines phenotypic, pedigree and genomic information. A national single-step approach (SSP) for a comprehensive information (phenotype, pedigree and genotype markers) genetic evaluation was developed for final score of US Holsteins. Data included final scores recorded from 1955 to 2009 for 6,232,548 Holsteins cows. BovineSNP50 genotypes from the Cooperative Dairy DNA Repository were available for 6,508 bulls. Analyses used a repeatability animal model as is currently used for the national US evaluation. Analyses included pedigree and genomic-based relationships matrices. Full data sets and a subset of records (final scores up to 2004) were used to estimate the increase in accuracy due to genomic information. Also, comparisons include a multiple-step approach for genomic selection. The SSP genetic evaluation with the pedigree relationship matrix augmented with genomic information provided genomic predictions with accuracy and bias comparable to multiple-step procedures. The implementation of such SSP requires the inverse of a join relationship matrix based on pedigree and genomic relationships. A second study investigated efficient computing options for creating relationship matrices based on genomic markers and pedigree information as well as their inverses. A matrix of incidence of SNP marker information was simulated for a panel of 40K SNPs. The number of genotyped animals varied from 1,000 to 30,000. Efficient methods to create the matrices used in the unified approach are presented. Optimizations can be obtained either by modifications of the existing code or by the use of automatic optimizations provided by open source or third-party libraries. The third study evaluated the feasibility and accuracy of multiple trait evaluation for conception rate (CR) defined as outcomes of all inseminations in US Holsteins using all available phenotypic, pedigree and genomic information. Genetic evaluations used a national data set and a multiple trait model. The evaluations were obtained by regular BLUP or by the SSP, using genomic information. The R2 obtained with the SSP were almost doubled compared to BLUP. Computing with SSP took 33% more time than with BLUP. A multiple trait evaluation of CR using the genomic information is possible and advantageous.