Combining genetic test information and correlated phenotypic records for breeding value estimation

¹ Animal and Dairy Science Department, University of Georgia, Athens, Georgia 30602-2771
² Animal and Dairy Science Department, University of Georgia, Athens, Georgia 30602-2771; Department of Statistics, University of Georgia, Athens, Georgia 30602-2771; Institute of Bioinformatics, University of Georgia, Athens, Georgia 30602-2771

^* To whom correspondence should be addressed. E-mail: mspanky{at}uga.edu.

	Abstract

The use of marker assisted selection in the beef cattle industry to date has been comprised of using traditional expected progeny differences in tandem with molecular test information. In the current study a multiple trait simulation was carried out to create a beef cattle data set using genetic parameter estimates from the literature in order to identify the best procedure for combining both sources of information and to assess the added benefit of the procedure. To reach these objectives the following simulation/analysis steps were implemented: 1) varying percentages of available records for the trait of interest (100, 5, 0) 2) varying percentages of animals with molecular information (100, 50, 25, 0) 3) scenarios where the favorable allele was more frequent (F) and when the unfavorable allele was more frequent (U) and 4) analyze the response due to selection over five generations. The data sets included three correlated traits where two of them, birth weight and average post weaning gain, had complete recording and the availability of records for the third trait (marbling score) varied. It was further assumed that molecular information was available for the third trait for a causative gene that explained 10% of the genetic variation. Estimates of Pearson correlations between true and predicted breeding values for marbling score declined as the amount of information declined and instances where molecular information was recorded were always closer to the true values than in the case when molecular information was absent. When the unfavorable allele was more frequent, rank correlation estimates were increased among top sires, low accuracy sires, and high accuracy sires by approximately 24.9%, 12.1%, and 4.7% with limited marbling score records and complete genotyping as compared to limited marbling score records and no genotyping. Similar results were seen when the favorable allele was more frequent. When there was a complete absence of recording for the trait of interest the same trends in correlations were observed and were lower than when the trait of interest was recorded. Jointly considering molecular and phenotypic information showed greater long term response as compared to tandem selection, showing that discrimination of candidates for selection based solely on molecular information is not optimal.

Key Words: beef cattle, best linear unbiased prediction, gene-assisted selection