Alternativas para avaliação genética de características produtivas e reprodutivas em populações multirraciais de bovinos leiteiros

Abstract

In this thesis, two papers were developed with the objective of checking if random regression models with linear spline types (RRMLS) are suitable for obtaining genetic parameters for three distinct traits: age at first calving (AFC), the first lactation length (FLL), and milk production up to 305 days in the first lactation (MY305), in a multiracial population of dairy cattle. Another aim was to investigate the genetic group effect of the progeny in the breeding values of bulls. To adjust multiple-triat models (MTM), the same characteristic evaluated in genetic groups 1/2 Holstein + 1/2 Gyr (1/2HG), 5/8 Holstein + 3/8 Gyr (5/8HG) and 3/4 Holstein + 1/4 Gyr (3/4HG) was considered as distinct traits. With RRMLS models the knots were adjusted to genetic groups 1/2HG, 5/8HG and 3/4HG. The RRMLS has presented the best fit. The additive and residual variances estimated by MTM and RRMLS models were similar. The heritabilities have ranged from 0.20 to 0.33 (AFC), 0.09 to 0.22 (FLL) and 0.15 to 0.35 (MY305) according to the genetic composition of cows. In the second study, the breeding values were modeled by random regression with linear splines with three, five and seven knots, arranged according to Holstein fraction in each genetic group of progeny. The knots were placed in groups Gyr (G), 1/2HG, and 7/8 Holstein + 1/8 Gyr (7/8HG) (model with 3 knots); G, 1/4 Holstein+ 3/4 Gyr (1/4HG), 1/2HG, 3/4HG and 7/8HG (model with 5 knots), and G, 1/4HG, 3/8HG, 1/2HG, 5/8HG, 3/4HG and 7/8HG (model with 7 knots). The residual variances were considered homogeneous or heterogeneous. The model with seven knots, for adjusting mean trajectories and genetic effects, with homogeneous residual variance, best fitted for AFC and FLL. The model with five knots and homogeneous residual variance showed best the fit for MY305. The posteriori means for heritability have ranged from 0.21 to 0.48 (AFC), from 0.21 to 0.38 (FLL) the and from 0.10 to 0.33 (MY305). The correlation between the random regression coefficients was low to moderate magnitude, and, in general, the high density intervals had included the value zero. The random regression models with linear splines allow to model the genetic heteroscedasticity and to estimate specific parameters for each genetic group. The existence of interaction between the breeding values of the bulls and the genetic group of their progeny was demonstrated in both studies. There may be errors in the classification of animals if this interaction is not taken into consideration. The two strategies of analysis presented in this study allow us to evaluate and rank the sires according to the genetic group to be bred. Multiple-trait models should be used when there were few genetic groups of progenies, while random regression models with linear splines must be taken in cases when there were a great number of genetic groups with available data.