4 Heritability Estimations for Intramuscular Fat in Hereford Cattle Using Random Regressions.

Random regressions make genetic predictions and parameter estimates possible across environmental gradients, which might possibly allow more accurate identification and beneficial use of breeding animals in specific environments. The objective of this study was to use random regression models for the estimation of heritability of intramuscular fat (IMF) in Hereford cattle, across a longitudinal gradient in the United States. Records were obtained from the American Hereford Association (n = 169,440) that included pedigree information from 227,902 animals. Three models were evaluated using ASReml: a quadratic random regression, a linear random regression, and a model without random regression. For all models, the fixed component involved the effects of contemporary group and ecoregion, where ecoregions were defined based on temperature and humidity designations across the United States. The random component considered the random effect of the animal itself, or as interacting with a linear or quadratic regression, using the longitude coordinates where the animal was reared as the regressor variable. The fit of the models was evaluated through likelihood-ratio tests, where the quadratic regression model proved to have significant advantages in comparison to the rest (P < 0.01). Heritability estimates using the quadratic model ranged from 0.31 to 0.54, having maximum value at the western coordinate evaluated (124.09 degrees west). Then, advancing from west to east the IMF heritability decreased reaching its minimum value at 99 degrees west. Furthermore, the heritability began to increase again, reaching values around 0.47 at the eastern coordinate evaluated (71.47 degrees west). These results indicate that quadratic random regressions may make improved parameter estimations and therefore improved genetic predictions for IMF in American Hereford cattle. This information may help to generate more precise selection indexes across different sectors of the United States. It may be possible that estimate differences could also occur for other economically relevant traits and other breeds, and further research is needed for this phenomenon.