Genetic parameters for milk yield and lactation persistency using random regression models in Girolando cattle.
In pasture based milk production systems of the Brazilian tropical and subtropical regions there is a predominance of the Girolando breed, animals which have better adaptation to the particular climatic conditions of these regions. The breed was developed from the 1940s by crossing Gir and Holstein. Majority of the Brazilian milk production, corresponding to 80% of the country's total production, comes from animals of this breed. Genetic evaluations for milk yield in Girolando breed have been based on accumulated milk production throughout lactation for a 305-days milk yield period (305MY). To calculate 305MY, the projection factors are used and the accuracy of these calculations depends on the quality of these factors as well as the quality and quantity of milk yield records available during this period. Alternatively, methods that directly use the test-day milk yield (TDMY) (random regression models, RRM) have been proposed in recent years and may overcome the problems arising from the use of 305MY.
The accuracy of the analysis by RRM depends crucially on the definition of the mathematical function to model fixed and random effects that act on the TDMY. Among various functions, Legendre's polynomials have been the most used ones. The random effects are usually modeled with polynomials of the same order (Cobuci et al., 2006; Cobuci and Costa, 2012), however, evidences in favor of using lower order polynomials to model the additive genetic effect in comparison to the permanent environment effect have also been reported (Pool et al., 2000; Lopez-Romero and Carabano, 2003; Bignardi et al., 2009). Besides estimating genetic parameters along the lactation curve, the RRM also allow the prediction of breeding values for lactation persistency, as these values are estimated at any time during lactation period (Jamrozik et al., 1997). The lactation persistency is defined as the ability of the cow in maintaining the level of milk yield after reaching the peak lactation and it is related to the reduction of production costs (Cobuci et al., 2007). Several measures of persistency have been proposed (Jamrozik et al., 1997; Jakobsen et al., 2002; De Roos et al., 2004), but there is no consensus regarding the best one to describe this feature.
Recent studies on assessment of lactation persistency in tropical regions reported unfavorable genetic association between milk yield and persistency (Dorneles et al., 2009b; Cobuci et al., 2012; Khorshidie et al., 2012; Pereira et al., 2012). Despite many studies on genetic parameter estimation via RRM as well as evaluation of lactation persistency in Holstein cows have been carried out in Brazil, there are no studies on Girolando breed. Therefore, by assessing Girolando breed via RRM this study aimed to: i) estimate (co)variance components and genetic parameters for TDMY; ii) evaluate different measures of lactation persistency in order to identify the most appropriate measure for use in genetic evaluation; iii) estimate genetic trends for TDMY.
MATERIAL AND METHODS
The data used in this study are from the National Animal Science Archive of Dairy Cattle, provided by the Genetic Improvement Program for Girolando Cattle (PMGG) under management of Embrapa Dairy Cattle, in partnership with the Associacao Brasileira dos Criadores de Girolando (ABCG). From this database were extracted 74,023 TDMY records using the first lactations of 9,162 Girolando cows that calved from 2000 to 2011 and with age at first calving varying from 23 to 56 months.
Aiming to obtain greater data consistency, only data from cows that had their first TDMY record up to 45 days after calving or with consecutive interval between TDMY exceeding 45 days were included; cows having at least four TDMY records per lactation collected between 5 and 305 days-in-milk (DIM); and those cows with both parents known. Contemporary group was characterized by classes of herd-year-month (HYM) of control, and required a minimum of three cows per class; and with those cows that were daughters of sires possessing at least three daughters into two herds. The calving seasons were defined as rainy season (October to March) and dry season (April to September).
After data editing, remained 32,817 TDMY records from 4,056 first calving cows, belonging to six genetic groups (1/4H to 3/4G, 3/8H to 5/8G, 1/2H to 1/2G, 5/8H to 3/8G, 3/4H to 1/4G, 7/8H to 1/8G) as their racial composition of Holstein (H)-Gir (G), daughters of 276 sires collected from 118 herds between 2000 and 2011. Notes that in this editing data were excluded more than half cows (61.4%), however the proportion of cows (and records) excluded was similar between the genetic groups ranging from 56.9% (3/8H to 5/8G) to 66.0% (5/8H to 3/8G), where 3/8H to 5/8G, 1/2H to 1/2G and 5/8H to 3/8G groups highlighted by the number of animals through the years (Figure 1). The pedigree file used in the calculation of coefficients of the Wright's numerator relationship matrix contained a total of 8,571 animals. Table 1 provides additional details of data structure used in analyzes.
Distribution of the number of TDMY records and the corresponding daily milk yield over the lactation are showed in Figure 2. The daily milk yield followed a typical lactation curve, increasing from the beginning of lactation until full production around the 64th day with 17.49 kg of milk, corresponding to the lactation peak period; afterwards the output was gradually decreased until the end of lactation. The number of TDMY records showed similar trend in milk yield, however, with higher fluctuations along the lactation curve (Figure 2).
Initially, six RRM were tested in order to identify the one which best fit to the production records. Models ranged in the orders of the Legendre's polynomial used to adjust the additive genetic effects and permanent environmental effects. The general structure of the RRM was:
[MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII] (1)
where: [y.sub.ijkln] is the nth observation, recorded on the lactation day (t) of the animal (l) from genetic group (k), corresponding to the calving season (j), and control class of HYM (i); HYM the fixed effect of the nth control class of HYM (i = 1, ..., 3070); [EC.sub.j] the fixed effect of the jth calving season (j = 1 and 2); [GG.sub.k] the fixed effect of the kth cow genetic group (k = 1, ..., 6); b the fth regression coefficient for linear (f = 1) and quadratic (f = 2) effects of the cows' age [x.sub.l] at calving, in months; [[beta].sub.r] are the rth fixed regression coefficients, specific for modeling the average lactation curve of the population, [[alpha].sub.lr] and [[gamma].sub.lr] are the rth random regression coefficients that describe the trajectory of the additive genetic effects and permanent environmental effects of the lth cow, respectively; [[psi].sub.r](t) is the covariate of regression function according to lactation day (t); [K.sub.o] is the adjustment order for the fixed regression coefficients corresponding to the average lactation curve of the population ([K.sub.o] = 3); [K.sub.a] and [K.sub.pe] are the adjustment orders of the Legendre's polynomial for the genetic additive random effects and permanent environmental effects, respectively, which ranged from 3 to 5, and [[epsilon].sub.ijkln] is the residual random effect associated with [y.sub.ijkln].
The models are referred to as [LegK.sub.a], [K.sub.pe], where [K.sub.a] and [K.sub.pe] represent the orders of Legendre's polynomials adjusted for additive genetic effects and permanent environmental effects, respectively. For example, the model Leg 3, 4 denotes an analysis that adjusts Legendre's polynomials of third and fourth order to the additive genetic effects and permanent environmental effects, respectively.
The residual variance was considered heterogeneous throughout the lactation period, as during prior analysis changes in the residual variance were detected over the lactation period. Therefore, the following classes of residual variance were defined: 5 to 30, 31 to 180, 181 to 240, 241 to 260, 261 to 280, and 281 to 305 DIM.
The RRM were compared using the following statistical criteria: the Akaike's information criterion (AIC) = -2LogL+ 2p (Akaike, 1973); the Schwarz's Bayesian information criterion (BIC) = -2LogL+pLog(n) (Schwarz, 1978); the logarithm of the likelihood function (-2logL); the weighted residual variance, WRV = (1/TDIM) x [f.summation over(l=2)] [[sigma].sup.t.sub.ef] [NDIM.sub.f]; the degree of interdependence between the model parameters, given by the second term of the of the theoretical information measure of the model's complexity, Complexity = [C.sub.1]([SIGMA]) (Bozdogan, 2000); relative percentage of complexity reduction (RPCR) = [[C.sub.1]([[summation].sub.k])-[C.sub.1]([[summation].sub.kR])/[C.sub.1](Xfc)] x 100 (Bozdogan, 2000), where p is the number of model parameters; n is the total number of observations; TDIM is the total number of days in milk; f o total number of residue classes; NDIM number of days in milk existing in the lth class; and [[summation].sub.k] and [[summation].sub.kR] are the matrices of (co)variance and correlation of the model parameters; [C.sub.1]([summation]) = rank([summation]) Log[trace([summation])/rank([summation])]-Log([absolute value of [summation]]). Lower values associated to these criteria indicate better model fit, except for RPCR where higher percentages indicate better fit. The (co)variance components were estimated by restricted maximum likelihood method using the Wombat program (Meyer, 2007).
Using the RRM that best fit the data, identified by the comparison criteria described above, genetic parameters were obtained for nine measures of persistency in milk yield (PSO (Table 2) and milk yield up to 305 DIM, as well as the genetic and permanent environmental correlations between these persistency measures and between these measures and 305MY. Lower values of persistency measures indicate greater lactation persistency, except for PS5, where higher values indicate greater lactation persistency.
The breeding value for milk yield to the animal (i) on any day (t) in milk, for example, to the RRM Leg 3, 3, Leg 3, 4 and Leg 3, 5 ([K.sub.a] = 3) can be obtained as follows: [[??]'.sub.i] = [[[??]'.sub.i0] [[??]'.sub.i1] [[??]'.sub.i2]], the vector of estimative for specific random regression coefficients for animal (i), and [Z'.sub.t] = [[[psi].sub.0t] [[psi].sub.it] [[psi].sub.2t]] the vector of coefficients of the Legendre's polynomial evaluated on day (t), then, the estimated breeding value (EBV) of the animal (i) on day (t) is:
[EBV.sub.it] = [Z'.sub.t] [[??].sub.i] = [[K.sub.a]-1.summation.j-0] [[alpha].sub.ij][[psi].sub.j] (t) (2)
Therefore, the EBV to 305MY ([EBV.sub.305]) of the animal (i) is achieved by:
[MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII] (3)
where: [MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII] is the vector sum of the coefficients of the Legendre's polynomials, specific to milk yield over the full lactation.
The genetic trend for 305MY was calculated from the annual average of sires and cows EBV, according to the regression model:
[y.sub.d] = [b.sub.0] + [b.sub.1][X.sub.d] (4)
where: [y.sub.d] is the mean EBV of the dth year of birth; [X.sub.d] is the dth year of birth; and [b.sub.0] and [b.sub.1], are respectively, the intercept and the linear regression coefficient (genetic tendency).
RESULTS AND DISCUSSION
Results of the comparison criteria of quality in adjustment of the models indicated that the Leg5,5 model showed the best fit by AIC criterion and by the logarithm of the likelihood function (Table 3). According to the RPCR, the Leg4,5 model provided a better fit to the data. The BIC criterion was even stricter in choosing the model due to the parameterization, thus indicating the Leg3,5 as the best model to fit genetic and permanent environmental effects. However, Leg3,5; Leg4,5; and Leg5,5 models presented the best fitting for weighted residual variance criteria, since those models had low and similar values. The complexity degree tended to increase with the number of estimated parameters.
The disagreement between the fit quality criteria when choosing the best model, observed in this study, has also been reported by other authors (Lopez-Romero and Carabano, 2003; Pelicioni et al., 2009). In general, the comparison criteria used in this study indicated that Leg3,5 and Leg5,5 models showed the best fit to the data. On analysis, it was observed that the estimates of the residual variance and genetic variance of both models were similar (Figure 3a and 3b), indicating that increasing the order of the Legendre's polynomial for modeling additive genetic effect causes little distinction in heritability estimates between the models (Figure 3c). There was also a greater variation in these estimates between the DIM throughout the lactation period when used, in the Leg5,5 model, the Legendre's polynomial of order 5 to modeling this additive genetic effect, suggesting that it is a less robust model. Moreover, the complexity of the Leg5,5 model was higher than that of Leg3,5 model (Table 3). Therefore, it is recommended the Leg3,5 model to describe the changes in the (co)variance components in milk yield during lactation of Girolando cows.
Results from this study are similar to those reported by Lopez-Romero and Carabano (2003), who compared RRM in dairy Holstein cattle in three regions of Spain and observed differences between BIC, AIC, and-2logL, where the Leg3,5 model was elected by the BIC criterion as the best one, since it is the most plausible in specifying additive genetic effects and permanent environmental effects. Pool et al (2000) and Bignardi et al. (2009) also reported that the additive genetic effect modeling with lower order Legendre's polynomials for the permanent environmental effect are sufficient to capture most of the genetic and permanent environmental variability observed in the shape of lactation curve. However, other authors have reported better fits with Legendre's polynomials of the same order for both random effects (Cobuci et al., 2006).
When assessing the estimates of (co)variance additive genetic and permanent environmental components between the random regression coefficients it revealed that the variances of the first two coefficients of additive genetic effects were greater than the third one (Table 4), indicating that they have contributed significantly in the variation description of the TDMY records. Indeed, the eigenvalues of these two factors explained more than 98% of the total variability of these records. This value is similar to the one reported by Pool et al. (2000) for Holstein in Denmark, where the first two eigenvalues explained 97.2% of the total genetic variation by the Leg3,5 model. For the permanent environmental effect, four coefficients were necessary to explain this percentage of variation.
Heritability estimates, additive genetic, permanent environmental, residual and phenotypic variances along the lactation curve showed well-defined trends (Figure 4). It can be seen that the additive variance was slightly higher in early lactation and thereafter, became stable until the end of lactation. The permanent environmental and phenotypic variances presented similar trends, however greater than additive genetic variance, with higher values in early lactation, remaining stable in the middle and increasing at the end of lactation. Similar trends have been reported in studies with Holstein in Brazil (Bignardi et al., 2009; Dorneles et al., 2009; Biassus et al., 2011). The highest values observed at the extremes of the curve can be attributed to variations in the number of TDMY records, yield level, non-genetic factors that were not considered in the model and pregnancy effects (Bohmanova et al., 2008).
Heritability estimates were low in early lactation, with values close to 0.18 in the first 30 days, gradually increasing as lactation advances until reaching the maximum value on day 205 ([h.sup.2] = 0.23). From then, gradually decreased until the end of lactation with values of 0.18 on day 305 (Figure 4). Lower values observed in early lactation can be attributed to the higher permanent environmental variance and to the lower additive genetic variance in this period. Similar trend was observed by Dorneles et al. (2009a) for Holsteins in Brazil, with heritability estimates ranging from 0.14 to 0.20. In general, genetic correlations were higher between adjacent DIM and lower among the more distant ones (Table 5). DIM in mid lactation were highly correlated, with correlations greater or equal to 0.90 between DIM on day 155 with DIM from day 105 to 255 (Table 5 and Figure 5a). The permanent environmental correlations were also high between adjacent DIM, decreasing between the farthest ones (Figure 5b). It can also be observe low and negative genetic correlation between the extreme controls (5 and 305). Negative genetic correlations between extreme controls during lactation period have been reported in the literature (Lopez-Romero and Carabano, 2003; Costa et al., 2005; Bignardi et al., 2009; Pereira et al., 2010).
According to Lopez-Romero and Carabano (2003), RRM tends to overestimate genetic variances and heritability at the extremes of the lactation curve and underestimate the genetic correlations between the first and last DIM. In this study, the number of observations at the end of lactation was low, which could have contributed to the underestimation of the correlation between the extremes of the lactation curve (Figure 2). This fact was expected as it is natural that Girolando cattle exhibit short period of lactation, which is typical of Bos indicus animals.
The heritability estimates for persistency measures ranged from 0.10 to 0.33 and were higher in [PS.sub.3] and [PS.sub.7] measures, and lower in the [PS.sub.4], [PS.sub.5], [PS.sub.6], and [PS.sub.8] measures (Table 6), suggesting that there is an important contribution from the genetic component to this trait, which could be included among the criteria for selection in Girolando breed. Differences between heritability estimates of persistency measures may be associated to the period (phase) of lactation used in the calculation for persistency (Madsen, 1975). Cobuci et al. (2012), when working on data from Holstein using fourth-order polynomials, reported heritability estimate of 0.33 for PS3, equal to the obtained in the present study. In general, the values obtained in this study were higher than those reported by Pereira et al. (2012) for Gir, and Biassus et al. (2010) and Dorneles et al. (2009b) for the Holstein breed in Brazil, which ranged from 0.04 to 0.32 and were obtained with RRM in third and fourth-order Legendre's polynomials.
The heritability estimate for milk yield up to 305 days (0.27) lies within the range reported in studies with RRM using different polynomial functions (Dorneles et al., 2009b; Biassus et al., 2010; Pereira et al., 2012). The heritability estimate for milk yield up to 305 days obtained in this study is higher than the 0.21 and 0.25 obtained by Pereira et al. (2012) for Gir breed and by Dorneles et al. (2009b) for Holsteins, respectively. However, it is lower than the values obtained by Jakobsen et al. (2002), Cobuci et al. (2006) and Biassus et al. (2010), which ranged from 0.31 to 0.42 for Holsteins.
Estimates of genetic correlations between different persistency measures were of high magnitude, except between [PS.sub.5] with [PS.sub.2] and [PS.sub.4], and between [PS.sub.2] and [PS.sub.6] (Table 6). It can also be observed negative estimates between [PS.sub.5] and the other measures, probably due to the definition of [PS.sub.5], since higher values of this measure indicate greater persistency, unlike the other measures. In general, permanent environmental correlations between these measures were also high, except the correlations among [PS.sub.2] and [PS.sub.1], [PS.sub.6], [PS.sub.7] or [PS.sub.8]; [PS.sub.6] and [PS.sub.4] or [PS.sub.9]; and [PS.sub.5] and [PS.sub.9], which were low. In general, values from -0.92 to 1.00 among persistency measurements obtained in this study were similar to those reported by Pereira et al. (2012) for Gir (0.86 to 0.99), by Dorneles et al. (2009b) for Holstein (-0.87 to 0.94) and by Freitas et al. (2010) for Guzerat cattle (-0.99 to 0.97) in Brazil.
Genetic correlations among PS1, PS7, or PS8 and 305MY were close to zero, and close to those reported by Kistemaker (2003), who estimated genetic correlation of 0.06 between PS7 and 305MY. However, Dorneles et al. (2009b), Cobuci et al. (2012) and Khorshidie et al. (2012) studying the Holstein breed, and Jakobsen et al. (2002), studying the Danish breed, obtained values close to zero between PS5 and 305MY. These values confirm that there is a little genetic association between milk yield and persistency, indicating that animals having the same level of milk yield may present different levels of lactation persistency. Therefore, animals with greater EBV for persistency are not exactly those with greater EBV for 305MY (Cobuci et al., 2007). This fact can be confirmed when comparing the EBV for measures of persistency and 305MY from the top five sires with more than 25 daughters (Table 7), in which the fourth best sire (S4) classified according to its EBV to 305MY is the first one to all measures of persistency.
According to Jakobsen et al. (2002), the ideal measure for lactation persistency is the one which has expressive economic value, greater genetic variation, high heritability and low correlation with milk yield up to 305 days. In this study, the measure PS7 presented such characteristics therefore can be recommended for evaluation of lactation persistency in Girolando cattle.
The percentage of animals in common with more than 25 daughters selected for PS7 and 305MY is much lower at high intensities (<10%), however, as expected, this percentage increased as the proportion of selected animals was higher (Figure 6). It was also noted that this value is higher in cows than in sires, but when the sire selection is restricted by the number of daughters, this percentage becomes much higher. Similar trends were observed by Cobuci et al. (2007) for Holsteins and by Pereira et al. (2012) for Gir.
Estimated breeding values for milk yield during lactation of the top five sires for 305MY with more than 25 daughters showed similar patterns, however with marked differences as well (Figure 7). There were clear differences in predicted breeding values of sires throughout lactation. S1 and S4 sires showed the highest EBV at the lactation peak, however, in the end of period, the sires S2, S3, and S5 showed higher genetic merit for persistency. Likewise, Figure 8 presents the EBV from the top five sires ranked by lactation persistency ([PS.sub.7]). The EBV from sires S1 to S4 were higher than those from S5 and showed an increasing trend throughout lactation. Nevertheless, it is important to note that these five sires had different ratings for both traits (305MY and [PS.sub.7]).
Trends showed in Figures 7 and 8 indicate that the lactation persistency does not depend on the production level of animals, thus confirming the weak association between both traits (Cobuci et al., 2007; Pereira et al., 2012).
Estimates of annual genetic trend for 305MY for sires and cows were positive and of moderate magnitude, however negative for PS7 (Table 8). In general, the results revealed a significant genetic progress for 305MY in the Girolando population during the study period, a trend which was more pronounced in sires than in cows, with 31.73% increase over the study period. These results indicate efficiency of the breeding program for Girolando cattle.
The values of genetic trend obtained in this study are close to 6.71 and 7.53 kg of milk per year reported by Ferreira et al. (2006), however lower than 9.51 kg of milk per year obtained by Boligon et al. (2005) for Holsteins. Nevertheless, the annual genetic trends for lactation persistency for sires and cows were negative and of low magnitude (Table 8), indicating no genetic progress for this trait in the period 1997 through 2008. These results were expected, since the genetic correlation between 305MY and PS7 was close to zero (0.02, Table 6), confirming that the selection for 305MY does not lead to improvement in lactation persistency. Similar trends were observed by Cobuci et al. (2007) for Holsteins and by Pereira et al. (2012) for Gir. Based on these results, an alternative for the improvement of both traits implies in using selection indexes that combine lactation persistency and 305MY and enabling the simultaneous selection of the most productive and persistent animals for milk yield (Chaves, 2009). The study revealed that selection for increased to milk yield up to 305 days did not cause genetic progress for lactation persistency of the animals belonging to the Genetic Improvement Program for Girolando cattle. The use of the measure persistency PS7 proposed by Kistemaker (2003), under the RRM using 3 and 5-order Legendre's polynomials for additive genetic effects and permanent environmental effects, respectively, would be the most suitable option for use in genetic evaluation for milk yield and lactation persistency of the breed. The use of selection indexes for simultaneous evaluation of both genetic traits can be a viable alternative. However these are not the only trait to be considered in genetic evaluation, mainly in production systems of tropical regions.
The authors thank the Associacao Brasileira dos Criadores de Girolando for supplying the data.
Akaike, H. 1973. Information theory and an extension of the maximum likelihood principle. In: Proc. 2nd Int. Symp. Information Theory (Eds. B. N. Petrov and F. Csaki). Akademia Kiado, Budapest, Hungary. pp. 267-281.
Biassus, I. O., J. A. Cobuci, C. N. Costa, P. R. N. Rorato, J. Braccini Neto, and L. L. Cardoso. 2010. Persistence in milk, fat and protein production of primiparous Holstein cows by random regression models. Rev. Bras. Zootec. 39:2617-2624.
Biassus, I. O., J. A. Cobuci, C. N. Costa, P. R. N. Rorato, J. Braccini Neto, and L. L. Cardoso. 2011. Genetic parameters for production traits in primiparous Holstein cows estimated by random regression models. Rev. Bras. Zootec. 40:85-94.
Bignardi, A. B., L. El Faro, V L. Cardoso, P. F. Machado, and L. G Albuquerque. 2009. Random regression models to estimate test-day milk yield genetic parameters Holstein cows in Southeastern Brazil. Livest. Sci. 123:1-7.
Bohmanova, J., F. Miglior, J. Jamrozik, I. Misztal, and P. G. Sullivan. 2008. Comparison of random regression models with Legendre polynomials and linear splines for production traits and somatic cell score of Canadian Holstein cows. J. Dairy Sci. 91:3627-3638.
Boligon, A. A., P. R. N. Rorato, G B. B. Ferreira, T. Weber, C. J. Kippert, and J. Andreazza. 2005 Heritability and genetic trend for milk and fat yields in Holstein herds raised in the state of Rio Grande do Sul. Rev. Bras. Zootec. 34:1512-1518.
Bozdogan, H. 2000. Akaike's information criterion and recent developments in information complexity. J. Math. Psychol. 44:62-91.
Chaves, L. C. S. 2009. Genetic evaluation of dairy buffaloes using random regression model. Ph.D Thesis, Universidade Federal de Vicosa, Vicosa, MG, Brazil.
Costa, C. N., C. M. R. Melo, C. H. C. Machado, A. F. Freitas, I. U. Packer, and J. A. Cobuci. 2005. Estimation of genetic parameters for test day milk records of first lactation Gyr cows using repeatability and random regression animal models. Rev. Bras. Zootec. 34:1519-1530.
Cobuci, J. A., R. F. Euclydes, C. N. Costa, P. S. Lopes, R. A. Torres, and C. S. Pereira. 2004. Analysis of persistency in the lactation of Holstein cows using test-day yield and random regression model. Rev. Bras. Zootec. 33:546-554.
Cobuci, J. A., C. N. Costa, N. M. Teixeira, and A. F. Freitas. 2006. Use of Legendre polynomials and Wilmink function in genetic evaluations for persistency of lactation in Holstein cows. Arq. Bras. Med. Vet. Zootec. 58:614-623.
Cobuci, J. A., R. F. Euclydes, C. N. Costa, R. A. Torres, P. S. Lopes, and A. S. Pereira. 2007. Genetic evaluation for persistency of lactation in Holstein cows using a random regression model. Genet. Mol. Biol. 30:349-355.
Cobuci, J. A. and C. N. Costa. 2012. Persistency of lactation using random regression models and different fixed regression modeling approaches. Rev. Bras. Zootec. 41:1996-2004.
De Roos, A. P. W., A. G. F. Harbers, and G. De Jong. 2004. Random herd curves in a test-day model for milk, fat, and protein production of Dairy Cattle in the Netherlands. J. Dairy Sci. 87:2693-2701.
Dorneles, C. K. P., J. A. Cobuci, P. R. N. Rorato, T. Weber, J. S. Lopes, and H. N. Oliveira. 2009a. Estimation of genetic parameters for Holstein cows milk production by random regression. Arq. Bras. Med. Vet. Zootec. 61:407-412.
Dorneles, C. K. P., P. R. N. Rorato, J. A. Cobuci, J. S. Lopes, T. Weber and H. N. Oliveira. 2009b. Lactation persistency for Holstein cows raised in the State of Rio Grande do Sul using a random regression. Cienc. Rural 39:1485-1491.
Ferreira, W. J., N. M. Teixeira, R. A. Torres, and R. B. Barra. 2006. Estimate of genetic trend for milk production in Holstein in the State of Minas Gerais. Arq. Bras. Med. Vet. Zootec. 58:599-606.
Freitas, L. S., M. A. Silva, R. S. Verneque, B. D. Valente, G. S. Correa, R. F. Ferreira, M. G C. D. Peixoto, and G. G. Santos. 2010. Evaluation of lactation persistency of Guzerat cows using random regression models. Arq. Bras. Med. Vet. Zootec. 62:401-408.
Jakobsen, J. H., P. Madsen, J. Jensen, J. Pedersen, L. G. Christensen, and D. A. Sorensen. 2002. Genetic parameters for milk production and persistency for Danish Holsteins estimated in random regression models using REML J. Dairy Sci. 85:1607-1616.
Jamrozik, J., L. R. Schaeffer, and J. C. M. Dekkers. 1997. Genetic evaluation of dairy cattle using test day yields and random regression model. J. Dairy Sci. 80:1217-1226.
Khorshidie, R., A. A. Shadparvar, N. Ghavi Hossein-Zadeh, and S. Joezy Shakalgurabi. 2012. Genetic trends for 305-day milk yield and persistency in Iranian Holsteins. Livestock Science, 144:211-217.
Kistemaker, G. J. 2003. Comparison of persistency definitions in random regression test day models. In: Proceedings of the Interbull Technical Workshop. 2003 March, 1-3; Beltsville, MD, USA. Interbull Bulletin 30:96-98.
Lopez-Romero, P. and M. J. Carabano. 2003. Comparing alternative random regression models to analyse first lactation daily milk yield data in Holstein-Friesian cattle. Livest. Prod. Sci. 82:81-96.
Madsen, O. 1975. A comparison of some suggested measures of persistency of milk yield in dairy cows. Anim. Prod. 20:191-197.
Meyer, K. 2007. WOMBAT: a tool for mixed model analyses in quantitative genetics by restricted maximum likelihood (REML). J. Zhejiang Univ. Sci. B, 8:815-821.
Pelicioni, L. C., L. G. Albuquerque, and A. S. Queiroz. 2009. Estimates of covariance components for body weights from birth to 365 days of age in Guzera cattle, using random regression models. Rev. Bras. Zootec. 38:50-60.
Pereira, R. J., P S. Lopes, R. S. Verneque, M. L. Santana Junior, M. R. Lagrotta, and R. A. Torres. 2010. Covariance functions for test-day milk yield in Gir cattle. Pesq. Agropec. Bras. 45: 1303-1311.
Pereira, R. J., R. S. Verneque, P. S. Lopes, M. L. Santana Junior, M. R. Lagrotta, R. A. Torres, A. E. Vercesi Filho, and M. A. Machado. 2012. Milk yield persistency in Brazilian Gyr cattle based on a random regression model. Genet. Mol. Res. 11:1599-1609.
Pool, M. H., L. L. G. Janss, and T. H. E. Meuwissen. 2000. Genetic parameters of legendre polynomials for first parity lactation curves. J. Dairy Sci. 83:2640-2649.
Schwarz, G. 1978. Estimating the dimension of a model. Ann. Statist. 6:461-464.
Ali William Canaza-Cayo (1), Paulo Savio Lopes (2), Marcos Vinicius Gualberto Barbosa da Silva (3), Robledo de Almeida Torres (2), Marta Fonseca Martins (3), Wagner Antonio Arbex (3), and Jaime Araujo Cobuci (*)
Animal Science Department, Universidade Federal do Rio Grande do Sul, Porto Alegre, RS 91540-000, Brazil
* Corresponding Author: Jaime Araujo Cobuci. Tel: +55-5133087421, E-mail: email@example.com
(1) Facultad de Ciencias Agrarias, Universidad Nacional del Altiplano, Puno 21001, Peru.
(2) Animal Science Departament, Universidade Federal de Vicosa, Vicosa, MG 36570-000, Brazil.
(3) Embrapa Gado de Leite, Eugenio do Nascimento Street, 610, Juiz de Fora, MG 36038-330, Brazil.
Submitted Aug. 15, 2014; Revised Dec. 2, 2014; Accepted Mar. 24, 2015
Table 1. Descriptive statistics of the analyzed data N Average SD Number of TDMY 32,817 -- -- Number of herds 118 -- -- Number of contemporary groups 3,070 -- -- Number of lactations 4,056 -- -- Number of TDMY per lactation -- 8.09 1.94 Daily milk yield (kg) -- 15.10 4.47 Age at calving (d) -- 1,085.11 177.27 SD, standard deviation; TDMY, test-day milk yield. Table 2. Measurements of persistency in milk yield Measurements Author P[S.sub.1] = (EB[V.sub.280] - Jamrozik et al. (1997) EB[V.sub.60]) P[S.sub.2] = ([205.summation Jakobsen et al. (2002) over (t=106)] EB[V.sub.t] - [105.summation over (t=6)] EB[V.sub.t]) P[S.sub.3] = ([305.summation Jakobsen et al. (2002) over (t=106)] EB[V.sub.t] - [105.summation over (t=6)] EB[V.sub.t]) P[S.sub.4] = [280.summation Jamrozik et al. (1997) over (t=106)] (EB[V.sub.t] - EB[V.sub.60]) P[S.sub.5] = ([279.summation Jakobsen et al. (2002) over (t=60)] (EB[V.sub.t] - EB[V.sub.280]) P[S.sub.6] = (EB[V.sub.290] - Cobuci et al. (2004) EB[V.sub.90]) P[S.sub.7] = ([1/51] Kistemaker (2003) [305.summation over (t=255)] EB[V.sub.t] - [1/21] [70.summation over (t=50)] EB[V.sub.t]) P[S.sub.8] = ([300.summation Poso (2003) cited by over (t=101)] EB[V.sub.t] - Kistemaker (2003) 200 * EB[V.sub.100]) P[S.sub.9] = ([305.summation De Roos et al. (2004) over (t=61)] EB[V.sub.t] - 245 * EB[V.sub.60]) EBV, estimated breeding value; P[S.sub.I], measures of persistency in milk yield. Table 3. Number of estimated parameters (p), the logarithm of the likelihood function (-2logL), Akaike's information criterion (AIC), Bayesian information criterion (BIC), weighted residual variance (WRV), relative percentage of complexity reduction (RPCR) and degree of interdependence between the model parameters (complexity) for the models using different orders of orthogonal Legendre's polynomials Models (1) p -2LogL AIC BIC WRV RPCR Complexity Leg3,3 18 98,996 99,032 99,181 4.64 80.74 0.995 Leg3,4 22 98,714 98,758 98,940 4.29 84.06 1.312 Leg3,5 27 98,503 98,557 98,781 4.04 86.65 1.582 Leg4,4 26 98,710 98,762 98,978 4.29 86.58 1.915 Leg4,5 31 98,500 98,562 98,820 4.04 88.44 2.232 Leg5,5 36 98,432 98,504 98,802 4.04 84.76 2.839 (1) Leg3,3 = Leg[K.sub.a], [K.sub.pe], where [K.sub.a] and [K.sub.pe] represent the orders of Legendre's polynomials adjusted for additive genetic effects and permanent environmental effects, respectively. Table 4. Estimates of variances (diagonal) and (co)variances (below diagonal) between the random regression coefficients (1) with standard deviations in parentheses and eigenvalues ([lambda]) of (co)variances matrix of the additive genetic effects and permanent environmental effects with percentage of variance explained in brackets in Leg3,5 model (2) Coefficients of additive genetic effect [[alpha]. [[alpha]. [[alpha]. sub.0] sub.1] sub.2] [[alpha].sub.0] 5.331 (1.06) [[alpha].sub.1] 0.278 (0.40) 1.285 (0.28) [[alpha].sub.2] -0.587 (0.23) -0.221 (0.12) 0.190 (0.09) [[lambda].sub. 5.42 (79.6) 1.29 (19.0) 0.09 (1.38) [alpha]i.sup.3] Coefficients of permanent environmental effect [p.sub.0] [p.sub.1] [p.sub.2] [p.sub.0] 14.182 (0.93) [p.sub.1] 0.007 (0.35) 2.402 (0.25) [p.sub.2] -0.722 (0.23) -0.053 (0.12) 1.365 (0.11) [p.sub.3] 0.197 (0.11) -0.394 (0.06) -0.062 (0.05) [p.sub.4] 0.319 (0.11) -0.160 (0.06) -0.192 (0.04) [p.sub.5] 14.23 (74.85) 2.50 (13.15) 1.38 (7.25) Coefficients of permanent environmental effect [p.sub.3] [p.sub.4] p.sub.0] p.sub.1] p.sub.2] p.sub.3] 0.610 (0.05) p.sub.4] 0.095 (0.03) 0.454 (0.05) p.sub.5] 0.53 (2.81) 0.37 (1.94) (1) [[alpha].sub.i] and [p.sub.i] are the random regression coefficients for additive genetic effect and permanent environmental effect for the ith Legendre's polynomial, respectively. (2) Leg3,5, adopted model with three coefficients for the additive genetic effect and five coefficients for permanent environmental effect. (3) Eigenvalues of the (co)variances matrix of the regression coefficients of random effects. Table 5. Estimates of genetic correlations (above the diagonal) and permanent environmental correlations (below diagonal) between days in milk (DIM) selected from first lactation of Girolando cows DIM 5 30 55 80 105 130 155 180 5 0.98 0.90 0.79 0.67 0.55 0.44 0.34 30 0.85 0.97 0.91 0.81 0.72 0.62 0.53 55 0.67 0.95 0.98 0.92 0.85 0.78 0.71 80 0.57 0.87 0.97 0.98 0.94 0.89 0.84 105 0.51 0.78 0.90 0.97 0.99 0.96 0.92 130 0.48 0.69 0.81 0.91 0.98 0.99 0.97 155 0.45 0.62 0.73 0.84 0.93 0.98 0.99 180 0.42 0.58 0.68 0.78 0.87 0.93 0.98 205 0.38 0.55 0.64 0.71 0.78 0.84 0.91 0.97 230 0.34 0.54 0.60 0.64 0.67 0.71 0.79 0.88 255 0.32 0.53 0.58 0.59 0.59 0.61 0.68 0.78 280 0.34 0.53 0.57 0.57 0.57 0.57 0.62 0.70 305 0.38 0.43 0.45 0.48 0.51 0.52 0.53 0.53 DIM 205 230 255 280 305 5 0.25 0.18 0.10 0.04 -0.03 30 0.45 0.37 0.30 0.22 0.14 55 0.63 0.56 0.48 0.40 0.31 80 0.78 0.71 0.64 0.56 0.46 105 0.88 0.82 0.76 0.68 0.58 130 0.94 0.90 0.84 0.77 0.67 155 0.97 0.95 0.90 0.84 0.75 180 0.99 0.98 0.94 0.89 0.81 205 0.99 0.97 0.93 0.87 230 0.97 0.99 0.97 0.91 255 0.90 0.98 0.99 0.96 280 0.80 0.89 0.95 0.99 305 0.53 0.55 0.62 0.82 Table 6. Heritability estimates (diagonal), genetic correlations (above diagonal) and permanent environmental correlations (below diagonal) between measures of persistency ([PS.sub.1] to [PS.sub.9]) and milk yield up to 305 days (305MY) Trait [PS.sub.1] [PS.sub.2] [PS.sub.3] [PS.sub.4] [PS.sub.1] 0.18 0.88 0.99 0.95 [PS.sub.2] 0.50 0.28 0.94 0.98 [PS.sub.3] 0.96 0.65 0.33 0.98 [PS.sub.4] 0.76 0.92 0.85 0.15 [PS.sub.5] -0.80 0.10 -0.66 -0.23 [PS.sub.6] 0.94 0.20 0.82 0.52 [PS.sub.7] 1.00 0.51 0.95 0.77 [PS.sub.8] 0.93 0.49 0.93 0.79 [PS.sub.9] 0.77 0.87 0.85 0.95 305MY -0.07 0.06 -0.01 0.00 Trait [PS.sub.5] [PS.sub.6] [PS.sub.7] [PS.sub.8] [PS.sub.1] -0.92 0.98 1.00 1.00 [PS.sub.2] -0.64 0.77 0.88 0.90 [PS.sub.3] -0.86 0.94 0.99 0.99 [PS.sub.4] -0.76 0.86 0.95 0.96 [PS.sub.5] 0.10 -0.98 -0.93 -0.91 [PS.sub.6] -0.94 0.15 0.98 0.97 [PS.sub.7] -0.79 0.94 0.31 1.00 [PS.sub.8] -0.68 0.84 0.91 0.10 [PS.sub.9] -0.28 0.55 0.77 0.77 305MY 0.10 -0.12 -0.08 -0.07 Trait [PS.sub.9] 305MY [PS.sub.1] 0.97 0.02 [PS.sub.2] 0.97 0.30 [PS.sub.3] 0.99 0.10 [PS.sub.4] 1.00 0.21 [PS.sub.5] -0.80 0.21 [PS.sub.6] 0.89 -0.11 [PS.sub.7] 0.97 0.02 [PS.sub.8] 0.98 0.04 [PS.sub.9] 0.18 0.17 305MY 0.29 0.27 Table 7. Estimated breeding values (EBV) for measures of persistency ([PS.sub.1] to [PS.sub.9]) and milk yield up to 305 days (305MY) from the top five sires with more than 25 daughters ranked according to their EBV for milk yield Sire 305MY [PS.sub.1] [PS.sub.2] [PS.sub.3] [PS.sub.4] S1 1,043.72 -0.53 3,645.17 -147.28 9,356.66 S2 951.01 -1.36 9,291.61 -375.42 23,850.31 S3 939.69 -1.03 7,014.49 -283.41 18,005.25 S4 894.89 -0.05 317.17 -12.81 814.12 S5 863.49 -0.98 6,663.35 -269.23 17,103.94 Sire [PS.sub.5] [PS.sub.6] [PS.sub.7] [PS.sub.8] [PS.sub.9] S1 -22,412.46 -1.03 -0.55 -3,520.60 8,805.03 S2 -57,129.77 -2.63 -1.40 -8,974.07 22,444.19 S3 -43,128.82 -1.98 -1.06 -6,774.77 16,943.73 S4 -1,950.10 -0.09 -0.05 -306.33 766.12 S5 -40,969.85 -1.88 -1.00 -6,435.64 16,095.55 Table 8. Estimates of annual genetic trend (b), standard errors (SE) and coefficient of determination ([R.sup.2]) for milk yield up to 305 (305MY) and lactation persistency ([PS.sub.7]) of sires and cows Trait Sires Cows b SE [R.sup.2] b SE [R.sup.2] 305MY 8.5168 ** 2.31 0.58 6.4653 * 2.09 0.49 [PS.sup.7] -0.0042 * 0.00 0.37 -0.0039 * 0.00 0.41 * p<0.05, ** p<0.01.
|Printer friendly Cite/link Email Feedback|
|Author:||Canaza-Cayo, Ali William; Lopes, Paulo Savio; da Silva, Marcos Vinicius Gualberto Barbosa; Torres, R|
|Publication:||Asian - Australasian Journal of Animal Sciences|
|Article Type:||Author abstract|
|Date:||Oct 1, 2015|
|Previous Article:||Novel nucleotide variations, haplotypes structure and associations with growth related traits of goat AT motif-binding factor (ATBF1) gene.|
|Next Article:||Effect of the ratio of non-fibrous carbohydrates to neutral detergent fiber and protein structure on intake, digestibility, rumen fermentation, and...|