Multibreed Genomic Evaluation Using Purebred Dairy Cattle K

Multibreed Genomic Evaluation Using Purebred Dairy Cattle K. M. 1 Olson* and P. M. 2 Van. Raden 1 Department of Dairy Science Virginia Polytechnic and State University, Blacksburg, VA 20460 2 Animal Improvement Programs Laboratory, Agricultural Research Service, USDA, Beltsville, MD 20705 -2350 Number of Animals Genotyped by Breed INTRODUCTION The accuracy of genomic evaluations depends on the number of genotyped animals in the training data sets (data set used to make predictions), which is much lower for the breeds with smaller populations of dairy cattle in the U. S. Pooling genotyped animals from different breeds may increase the accuracy of the genomic predictions for all breeds. Traditional genetic evaluations use multibreed methods in dairy cattle evaluations, however, only within breed methods are currently used for genomic evaluations in the U. S. A common set of 43, 385 single nucleotide polymorphism (SNP) from over 30, 000 genotyped animals have made multibreed genomic evaluations possible. OBJECTIVE To investigate multibreed genomic methods using genotypes from Holstein, Jersey, and Brown Swiss dairy cattle. METHODS • Genotyped cows and bulls from the Holstein, Jersey, and Brown Swiss breeds were used • Method 1 estimated SNP effects separately within each breed and then applied the effects to another breed • Method 2 (across-breed) used a common set of SNP effects estimated from combined genotypes and phenotypes of all breeds Breed/ Class Training Validation Total Holstein Jersey 5, 331 2, 477 7, 808 1, 361 410 1, 771 RESULTS Brown Swiss 506 182 688 • Training animals had daughter or their own information as of Nov. , 2004 and were used to estimate the SNP effects • Validation animals had no daughter or own information as of Nov. , 2004, but had daughter / own information as of June, 2009 and were used to test the accuracy of the results • Results for protein yield were used to illustrate the comparison of the three different methods of using all breeds for genomic evaluation • Method 1 did not improve the predictive ability of genomic evaluations for protein yield over the traditional genomic method • Method 2 improved the accuracy for all three breeds and yielded the most favorable results (of all three methods) for the Brown Swiss and Holsteins • Method 2 across-breed GPTA accounted for less variability than the traditional GPTA for both the Holsteins and Jerseys • Method 3 increased the accuracy for all three breeds and yielded the best results for the Jerseys • Method 3 multi-breed GPTA accounted for more variability in the model than the traditional GPTA for all of the breeds Results - P-Values and coefficient of determination (R 2) for Traditional and Method 1 for protein yield Method 1 Breed Traditional Methods Holstein GPTA Jersey GPTA Brown Swiss GPTA < 0. 001 0. 873 0. 813 Jersey 0. 668 < 0. 001 Brown Swiss 0. 342 0. 844 Holstein R 2 Adjusted GPTA R 2 Adjusted < 0. 001 0. 5041 < 0. 001 0. 5045 0. 473 < 0. 001 0. 4858 < 0. 001 0. 4874 0. 107 0. 054 0. 0978 0. 081 0. 055 0. 1030 • Predicted transmitting ability (PTA) which is the traditional genetic evaluation method based on pedigree and phenotypes • Traditional genomic predicted transmitting ability (GPTA) – calculated within breed • Using another breed's SNP effects (method 1) did not help predict future performance of animals from different breeds • The all-breed and multi-breed (methods 2 & 3) models showed increases in the coefficient of determination • The increase in adjusted R squared were small (up to 3%) • The smaller breeds (ones with fewer observations) gained the most from using all the breeds in a genomic evaluation • Method 3 (multi-breed) used correlated SNP effects within breed estimated jointly using multitrait method • Multiple regressions to predict daughter deviations of protein yield were used to test the methods with the other effects in the model including: CONCLUSIONS Results – P-Values and coefficient of determination (R 2) for Method 2 and Method 3 for protein yield Method 2 Method 3 Breed Across Breed GPTA PTA R 2 Multi Breed Adjusted GPTA PTA R 2 Adjusted Holstein 0. 002 < 0. 001 0. 5063 < 0. 001 0. 742 < 0. 001 0. 5060 Jersey 0. 290 < 0. 001 0. 4876 < 0. 001 0. 324 < 0. 001 0. 4916 Brown Swiss 0. 007 0. 316 0. 088 0. 1337 0. 080 0. 140 0. 060 0. 1127 • The use of method 2 or method 3 maybe more important in crossbreed populations IMPACTS • The small gains in accuracy may not warrant the increased computational demands of a multibreed genomic evaluation based on the current U. S. dairy populations