Heritability Estimation and Modeling Strategies Utilizing Random Regression for Weaning Weight in American Hereford Cattle
Abstract
Sufficiently accounting for genotype-environment interactions would enhance genetic merit prediction and parameter estimation for traits of economic prominence in beef cattle. The objective of this work was to utilize random regressions to estimate additive direct genetic variation, and milking ability intercepts expressed as maternal additive genetic, or maternal permanent environmental variances, each as proportions of total of phenotypic variance for weaning weight, denoted as h^2, m^2, and c^2 , respectively, across latitude or longitude coordinates within the continental United States. Records came from the American Hereford Association (n = 226,845), with pedigree of 383,426 animals. Analyses were conducted across the entire continental United States across latitude or longitude. An animal model, linear random regression, and quadratic random regression model were employed with latitude as the random covariate. Both an animal model, and linear random regression model were utilized with longitude as random covariate. From the animal model, direct, maternal, and maternal permanent environmental variances as proportions of phenotypic variance for latitude and longitude were 0.19 ± 0.008, 0.04 ± 0.011, and 0.17 ± 0.011; and 0.19 ± 0.008, 0.04 ± 0.011, and 0.17 ± 0.011, respectively. For linear random regression across latitude or longitude, estimates of h^2 , m^2 , and c^2 were 0.09 to 0.20, 0.07 ± 0.010, and 0.14 ± 0.011; and 0.15 to 0.26, 0.03 ± 0.009, and 0.15 ± 0.010, respectively. For the quadratic random regression across latitude, estimates of h^2 , m^2 , and c^2 were 0.08 to 0.23, 0.07 ± 0.010, and 0.14 ± 0.010, respectively. The quadratic random regression across latitude was the best fit model (P < 0.001) for the data.
A quadratic random regression analysis across longitude was attempted, but failed to converge. Linear random regressions unique to regional subdivisions were attempted for latitude and longitude, but analyses failed to converge. Considerable differences in heritability estimates were observed in different geographical zones as modeled by random regressions across the entire continental United States, when modeling intercepts of genetic and environmental components of maternal milking ability, across latitude or longitude.
Citation
Langdon, John Michael (2019). Heritability Estimation and Modeling Strategies Utilizing Random Regression for Weaning Weight in American Hereford Cattle. Doctoral dissertation, Texas A&M University. Available electronically from https : / /hdl .handle .net /1969 .1 /189089.