Search

This study aimed to verify if random regression models using linear splines (RRMLS) are suitable for identifying genetic parameters in multiple-breed populations and also to investigate whether an interaction exists between the breeding value (BV) of sires and their progeny breed group. Ten populations were simulated by crossing 2 breeds with distinct genetic variance and nonzero segregation variance. To obtain the genetic parameters, 2 models were used: a multiple-trait model (MULT), in which the trait was considered distinct when evaluated in each group (1/2P1 + 1/2P2, 5/8P1 + 3/8P2, and 3/4P1 + 1/4P2), and a RRMLS with the spline polynomial knots adjusted to these same groups. The genetic parameters estimated through MULT and RRMLS did not differ from the simulated values. The correlations between BV (simulated and estimated) of animals were high and varied from 0.74 to 0.76, which indicates the efficiency of using MULT and RRMLS for predicting BV. Using field data, the traits age at first calving (AFC), first lactation length (LL), and 305-d milk yield (MY-305) from a multiple-breed population of Holstein-Gyr cattle were analyzed. The BV of animals were modeled through RRMLS with 3, 5, and 7 knots, distributed in accordance with the fraction of Holstein breed in each progeny breed group. It was verified that RRMLS with 7 knots for adjusting mean trajectories and genetic effects, with homogeneous residual variance, best fit AFC and LL. For MY-305, the best fit for mean trajectory and genetic effects was the RRMLS with 5 knots and with homogeneous residual variance. The posterior means of heritability varied from 0.21 to 0.48, 0.21 to 0.38, and 0.10 to 0.33 for AFC, LL, and MY-305, respectively. Estimates from genetic parameters obtained by using RRMLS with field data showed that this model is a useful tool for genetic evaluations of populations formed by a great number of breed groups. An interaction occurred between the BV of sires and their progeny breed group, and the genetic parameters for AFC, LL, and MY-305 traits from a multiple-breed population depend on breed composition of the progeny from which the evaluations are based. [ABSTRACT FROM AUTHOR]

• Different statistical models for analyzing genetic parameters for longitudinal traits. • Simulation processes to validate field data analyzed by single trait model. • Sampling records influence over estimation of genetic parameters. The comprehensive analyses of longitudinal traits under sequential selection could improve genetic parameters estimates and lead to more accurate selection decisions. The objective of this study was to evaluate statistical models for analyzing longitudinal traits under sequential selection. We used single trait (STM), multiple trait (MTM) and random regression model with linear splines polynomials (RRM) to estimate genetic parameters for body weight records of Nellore young bulls. First, we used a complete dataset (DS100) with 60,550 body weight records of 12,110 young bulls. Two additional datasets were also obtained from DS100. They were obtained with a sequential selection of 85% (DS85) and 70% (DS70) of heaviest animals. In addition, some datasets with the same number of records as DS85 and DS70 were also obtained with random sampling of 85% (RS85) and 70% (RS70) of body weight records at each age. Body weights were standardized at 330, 385, 440, 495 and 550 days of age for STM and MTM analysis. In RRM, the knots of linear splines were fitted at 250, 330, 385, 440, 495, 550 and 597 days of age. The estimates of additive genetic, residual and phenotypic variances from STM analysis of DS85 and DS70 were lower than the corresponding estimates from STM analysis of DS100. However, the estimates of genetic and environmental parameters from MTM and RRM analysis of DS100, DS85 and DS70 were similar. The reduction of dataset size with random sampling (RS85 and RS70) did not affect the estimates of genetic and environmental parameters from STM, MTM and RRM analysis. MTM and RRM are adequate for genetic evaluation of the longitudinal traits under sequential selection, but RRM presents some advantages over MTM. RRM with linear splines does not need previous adjustments of the body weights for standard ages and it also provides estimates of genetic and environmental parameters directly at the same points as the corresponding traits in MTM. [ABSTRACT FROM AUTHOR]