|Parker Gaddis, K|
Submitted to: Journal of Dairy Science
Publication Type: Abstract Only
Publication Acceptance Date: 2/23/2013
Publication Date: 7/8/2013
Citation: Parker Gaddis, K.L., Cole, J.B., Clay, J.S., Maltecca, C. 2013. Analysis of health trait data from on-farm computer systems in the U.S. II: Comparison of genomic analyses including two-stage and single-step methods. Journal of Dairy Science. 96(E-Suppl. 1):444 (abstr. 447). Interpretive Summary:
Technical Abstract: The development of genomic selection methodology, with accompanying substantial gains in reliability for low-heritability traits, may dramatically improve the feasibility of genetic improvement of dairy cow health. Many methods for genomic analysis have now been developed, including the “Bayesian Alphabet” and single-step methods. However, little research has been conducted to analyze the performance of these genomic methods when applied to lowly heritable traits, such as health events. This may be due in part to a lack of documented phenotypes for health events. Producer-recorded health information may be able to fill this gap and provide health-related phenotypes, allowing substantial improvements to be made in these traits. The principal objective of this study was to investigate various genomic methods applied to health data collected from on-farm computer systems in the US. A single-step analysis was conducted to estimate variance components and heritabilities of health traits commonly experienced by dairy cows including displaced abomasum, ketosis, lameness, mastitis, metritis, and retained placenta. A blended H-matrix was constructed by combining information from both the pedigree-based A-matrix and the marker-based G-matrix using sire information. A threshold model was used with fixed effects of parity and year-season and random effects of herd-year and sire. Two-step Bayesian methods were also implemented including single-trait Bayes A, as well as multiple-trait Bayes A analyses using deregressed sire breeding values as pseudo-phenotypes. The deregressed breeding values were obtained from previous analyses using threshold sire models. The data were split into four groups for cross-validation using K-means clustering based on relationship. Mean reliabilities of genomic estimated breeding values calculated with the single-step method ranged from 0.35 to 0.41. Mean reliability of genomic estimated breeding values, calculated using single-step methods, increased approximately 33% from previous estimates calculated from pedigree information for all traits. Comparable increases in reliability were obtained using two-stage methods. It was concluded that the addition of genomic information can improve the estimates of lowly heritable health traits.