Print version ISSN 1415-4757
Genet. Mol. Biol. vol.22 n.3 São Paulo Sept. 1999
S.G. Machado, M.A.R. Freitas and C.H. Gadini
Estação Experimental de Zootecnia de Ribeirão Preto IZ/CPA/SAA-SP, Av. Bandeirantes, 2419, 14030-670 Ribeirão Preto, SP, Brasil. Send correspondence to M.A.R.F.
Data were obtained from 17,968 records from 2,130 first lactations of Holstein cows calving between 1988 and 1991. The subjects were daughters of 136 sires monitored by Brazilian Breeders Association, Animal Science Institute, Department of Agriculture, a branch of the State of São Paulo. Data were divided into 10 subsets based on the number of days in milk yield. Test day milk yields (M1 to M10) and 305-day milk yield (M305) were the traits studied. These traits were adjusted for several environmental effects: class of cow age at calving, interval from calving to first test day, and herd-year-season. Restricted maximum likelihood estimates of (co)variance components were obtained from one and two-traits analysis under a sire model. Estimates of heritabilities for M ranged from 0.04 to 0.32. The highest values were found in the second half of lactation (M5 to M7). Heritability estimate for M305 was 0.32. Genetic correlations between individual test days and M305 ranged from 0.78 to 1.00. Results suggested that test day milk yields, mainly in mid-lactation, can be used instead of 305-day milk yield in genetic evaluations, because estimates of these two-trait heritabilities are nearly alike. Moreover, early selection can reduce generation intervals.
In general, standardized 305-day (d) milk yield (M305) has been used in genetic evaluation of dairy sires. Heifers that did not complete first lactation or had lactation periods shorter than 305 days have had their lactations extended and then used in the prediction of genetic merit.
However, prediction of uncompleted lactations may be biased. The major cause of bias in sire evaluations seems to be related to culling of heifers before completion of their first lactation, instead of non-random matings (Fimland, 1983).
To avoid extending lactations to 305-day, individual test day milk yields can be used in genetic evaluation of dairy sires and cows. Models that include test day milk yields in statistical analysis provide an alternative to selection on 305-day milk yield (Swalve, 1995b). In these models, any information on production can be utilized, and the models become more flexible. Cows can be included in the evaluations if they have at least one test day yield, and sires can be evaluated more accurately based on a larger amount of data available from their daughters instead of only one lactation record per daughter (Ptak and Schaeffer, 1993). The increase in accuracy is due mainly to a better adjustment of test day yields.
In Brazil, several studies have been done on aspects of the lactation curve, including studies to model lactation curves (Madalena, et al. 1979; Queiroz, et al. 1991; Durães, et al. 1991; Freitas, et al. 1992; Mcmanus, et al. 1997). There is no report on test day milk yield in specialized dairy breeds, probably due to the difficulty of storing enough information about a large number of animals. Test day records usually are discarded after M305 is calculated. Increases in computer storage capacity would allow this kind of information to be used either in lactation curve studies or directly in genetic evaluations.
Genetic parameters for test day yields have been estimated by several authors (Van Vleck and Henderson, 1961; Keown and Van Vleck, 1971; Auran, 1976; Danell, 1982, Swalve, 1995b, Gadini, 1997). According to these authors, selection based on some test day records is as efficient as on all. This may reduce the number of tests in routine milk recording.
The present study aims to estimate components of variance and covariance for test day milk yields in first lactation of Holstein cows by the use of a multiple trait derivative-free restricted maximum likelihood procedure, evaluating the possibility of utilizing these test day records as a selection criterion.
MATERIAL AND METHODS
Data were obtained from first lactation Holstein cows calving between 1988 and 1991. The herds were part of the Dairy Recording Program developed by the Brazilian Breeders Association, in partnership with the Animal Science Institute, a branch of the São Paulo State Department of Agriculture. Only first lactation records from cows aged between 1.5 and 3.5 years at calving were considered in this study.
Test day milk yields, recorded at approximately monthly intervals throughout lactation (M1-M10), and milk yields up to 305-days (M305) were the traits under analysis. According to Pander et al. (1992), test day yield was defined as the sum of all weighings of milk in a 24-h period.
The first interval in days for the first test was defined as date of recorded test day minus date of calving. Records with a first interval less than four or greater than 45 days were eliminated for further analysis. Because test days do not always follow a regular interval of 30 days, interval classes of 30 days were created to make the 10 test days uniform. Test day records after 305 days were eliminated. Lactations with at least two test day records were used (Meyer et al., 1989), or heritabilities and correlations for the first two tests were taken from the analysis of records of cows with at least two valid tests. Lactation records were discarded when the sire had less than five daughters, or when all the daughters were in the same herd.
Finally, a data set (Table I) of 17,968 test day records from 2,130 first lactations of Holstein cows, sired by 136 bulls, was used for further analysis. The 305-d milk yield was computed according to the method proposed by the Ministry of Agriculture (MAARA, 1986).
Two calving seasons were considered: the rainy season, October through March, and the dry season, April through September. Herd-year-season subclasses with at least four observations were used in the analysis. Age at calving was split into five classes (1.5 to 2.1; 2.2 to 2.4; 2.5 to 2.7; 2.8 to 3.1, and 3.2 to 3.5 years). Four classes were defined for the calving to first test day interval (from day 4 to day 14; 15 to 24; 25 to 35, and 36 to 45). Data were analyzed under restricted maximum likelihood, with the software MTDFREML (Multiple Trait Derivative-Free Restricted Maximum Likelihood), by Boldman et al. (1995). A sire model was used:
y = Xb + Za + e
where y = vector of dependent variables (M1 to M10 and M305), b = vector of fixed effects (age class, herd-year-season, interval calving-first control class), X e Z = design matrices for fixed and random effects, a = vector of sire effects, e = vector of residual effects. E(y) = Xb; E(a) = E(e) = 0. Var(a) = G = G0 Ä It and Var(e) = R = R0 Ä In, where G0 = matrix of additive genetic covariance between the traits within sire, R0 = matrix of residual covariance between traits within animal, I = identity matrix of order n (number of records) or t (number of sires), Ä = direct product operator.
Preliminary analysis on data set from Table I indicated that there was no significant effect of class of calving to first test day interval on test day milk yields from the third test day on, so that this effect was considered only for M1 and M2. Convergence criterion was defined for simplex variance below 10-9.
RESULTS AND DISCUSSION
Mean milk yield has the typical form of a lactation curve with the highest value occurring at M3 (Table II). In general, standard deviations were similar across test day records except for M6 and M7.
Estimates of heritabilities and the genetic and phenotypic correlations between each M and M305 (Table III) followed trends in the literature, with higher values found in the second half of lactation, between M5 and M7 (Meyer et al., 1989; Pander et al. 1992). In most reports, estimates were similar to heritability for M305. These published results suggested that test day records for mid-lactation could be used instead of M305 in genetic evaluations of dairy animals, but in none of the studies mentioned the estimates of heritability exceed those obtained for M305. This early selection can reduce generation interval. For M9 and M10, heritabilities were lower than those presented by Meyer et al. (1989), Rekaya et al. (1995) and Danell (1982). However, Kathenbrink and Swalve (1993), cited by Swalve (1995b), found heritability estimates of 0.09 and 0.05 for the last two test day milk yields, respectively.
Considerable differences in heritability estimates are found in the literature, due to differences in populations, different methods of analysis and, especially, how a trait is defined. In a study where several definitions for test day milk yield were analyzed simultaneously, Swalve (1995a) noted that yields defined as the average yield in a standardized 30-day interval resulted in higher estimates of heritability than for a non-standardized interval. He indicate that the use of standardized intervals can remove some of the problems that arise when different intervals between test days are used, especially between calving and first test day. Selection of sires and dams also can affect estimates. Auran (1976) reported slightly larger estimates from data where cows were not culled.
Genetic correlations between M1 and M305 varied from 0.78 to 1.00. Also, higher values were found in mid-lactation. However, other reports (Keown and Van Vleck, 1971; Pander et al., 1992; Ribas et al., 1994) have found significantly lower values, with genetic correlations between M1 and M305 and between M10 and M305 equal to 1.0. Moreover, when MTDFREML is used in the estimation of genetic correlations, there is a tendency for the estimates to reach +1 or -1, and the convergence to reach a local maximum. This may indicate the need for a larger data set to obtain better estimates (Van Vleck, 1992).
Rekaya et al. (1995) found the highest genetic correlation between the first test and M305 (0.89), but in the second half of the lactation genetic correlations between test day milk yields and M305 tended to decrease. Estimates for genetic correlations were higher than those found for phenotypic correlations, which followed the same trend in the literature, where smaller values were found in first and last test days, and the largest were in mid-lactation.
Residual variances were significantly larger in the first three test days (Table IV). These results are in accordance with Meyer et al. (1989) and Gadini (1997). Meyer et al. (1989) used univariate models with age at the test and days in milk as covariables, and they used herd year season (HYS) and herd test date (HTD) as fixed effects to fit test day production traits. They observed that including HTD in the model considerably reduced residual variances. Unfortunately, in our study we could not include HTD as a covariable because of the small data set.
This study has shown that genetic correlations between test day milk yields and 305-day milk yield are high and positive, indicating that test day milk yields could be utilized in place of M305 in genetic evaluations of dairy animals. The highest estimates of heritability in mid-lactation suggested that these test day yields (M5-M7) could be used as the selection criteria, leading to a reduction in generation interval.
The authors thank the Conselho Nacional de Desenvolvimento Científico e Tecnológico (CNPq) and the Fundação de Amparo à Pesquisa do Estado de São Paulo (FAPESP) for their financial support, and ABC/IZ-SAA/SP for providing the data. S.G.M. and M.A.R.F. are recipients of CNPq and FAPESP fellowships. Publication supported by FAPESP.
No presente estudo foram utilizados 17.968 registros de produção de leite, referentes a 2130 primeiras lactações de vacas da raça Holandesa, paridas nos anos de 1988 a 1991, filhas de 136 touros e controladas pela Associação Brasileira de Criadores (ABC). Os dados foram distribuídos em dez sub-arquivos de acordo com o número do controle (M1 a M10). As características estudadas foram: produção de leite no dia do controle (M) e produção aos 305 dias de lactação (M305), as quais foram ajustadas para os seguintes fatores de variação: idade da vaca ao parto em classes, intervalo parto-primeiro controle e subclasses de rebanho-ano-estação de parto. Os componentes de (co)variância foram obtidos a partir de análises com duas características utilizando-se o método REML, sob um modelo de touro. As estimativas de herdabilidade para as M variaram de 0,04 a 0,32 sendo os maiores valores encontrados na segunda fase da lactação (M5 a M7), enquanto que para a M305 este valor foi 0,32. As correlações genéticas entre as M e a M305 variaram de 0,78 a 1,00. Os resultados sugerem que as M podem ser utilizadas em substituição à M305 nas avaliações genéticas dos animais, visto que as herdabilidades no meio da lactação foram semelhantes à encontrada para a M305. Além disso, a seleção antecipada pode levar a uma redução no intervalo de gerações.
Auran, T. (1976). Studies on monthly and cumulative monthly yield records. III. Estimates of genetic parameters. Acta Agric. Scand. 26: 3-9. [ Links ]
Boldman, K.G., Kriese, L.A., Van Vleck, L.D., Van Tassell, C.P. and Kachman, S.D. (1995). A Manual for Use of MTDFREML USDA-ARS. Clay Center, Lincoln, NE. [ Links ]
Danell, B. (1982). Studies on lactation yield and individual test day yields of Swedish dairy cows. II. Estimates of genetic and phenotypic parameters. Acta. Agric. Scand. 32: 83-91. [ Links ]
Durães, M.C., Teixeira, N.M. and Freitas, A.S. (1991). Curvas de lactação de vacas da raça Holandesa em confinamento total. Arq. Bras. Med. Vet. Zootec. 43: 447-458. [ Links ]
Fimland, E. (1983). Some properties of using the additive genetic relationship in the mixed model technique. Z. Tierz. Zuechtungsbiol. 100: 361-374. [ Links ]
Freitas, M.A.R., Oliveira, H.N., Cardoso, V.L. and Oliveira, A.A. (1992). Fatores que influenciam a forma da curva de lactação de vacas da raça Holandesa. XXIX Reunião Anual da Sociedade Brasileira de Zootecnia, Anais, Lavras, MG, pp. 256. [ Links ]
Gadini, C.H. (1997). Genetic evaluations of test day production traits and somatic cell scores. Doctoral thesis, University of Nebraska, Lincoln. [ Links ]
Keown, J.F. and Van Vleck, L.D. (1971). Selection on test day fat percentage and milk production. J. Dairy Sci. 54: 199-203. [ Links ]
MAARA - Ministério da Agricultura, Abastecimento e Reforma Agrária (1986). Procedimentos para obtenção de variáveis calculadas. DOU - Diário Oficial da União. Seção I, 15.10.86, pp. 15.532-15.535. [ Links ]
Macmanus, C., Guth, T.L.F. and Saueressig, M.G. (1997). Curvas de lactação em gado Holandês em confinamento total no DF. Anais da XXXIV Reunião da Sociedade Brasileira de Zootecnia, Juiz de Fora, MG, Vol. 3, pp. 74-76. [ Links ]
Madalena, F.E., Martinez, M.L. and Freitas, A.S. (1979). Lactation curves of Holstein Friesian and Holstein Friezian x Gir cows. Anim. Prod. 29: 101-107. [ Links ]
Meyer, K., Graser, H. and Hammond, K. (1989). Estimates of genetic parameters for first lactation test day production of Australian Black and White cows. Livest. Prod. Sci. 21: 177-199. [ Links ]
Pander, B.L., Hill, W.G. and Thompson, R. (1992). Genetic parameters of test day records of British Holstein-Friesian heifers. Anim. Prod. 55: 11-21. [ Links ]
Ptak, E. and Schaeffer, L.R. (1993). Use of test day yields for genetic evaluation of dairy sires and cows. Livest. Prod. Sci. 34: 23-34. [ Links ]
Queiroz, S.A., Freitas, M.A.R., Albuquerque, L.G. and Lôbo, R.B. (1991). Fatores genéticos e de ambiente que influenciam os componentes da curva de lactação de bovinos da raça Holandesa. Arq. Med.Vet. Zootec. 43: 357-370. [ Links ]
Rekaya, R., Béjar, F., Carabaño, M.J. and Alenda, R. (1995). Genetic parameters for test day measurements in Spanish Holstein-Friesian. In: Proceedings of Interbull Annual Meeting. Prague, Czech Republic, pp. 1-8. [ Links ]
Ribas, M., Perez, B. and Guzman, G. (1994). Monthly test day milk yields and the extension of lactation in Cuba. Cuban J. Agric. Sci. 28: 125-139. [ Links ]
Swalve, H.H. (1995a). The effect of test day models on the estimation of genetic parameters and breeding values for dairy yield traits. J. Dairy Sci. 78: 925-938. [ Links ]
Swalve, H.H. (1995b). Test day models in the analysis of dairy production data - a review. Arch. Tierz. Dummerstorf. 38: 591-612. [ Links ]
Van Vleck, L.D. (1992). The resolution in statistical computing: from least squares to DFREML. In: Proceedings of National Breeders Roundtable, Poultry Breeders of America and South Eastern Poultry and Egg Association, St. Louis, MO, Vol. 41, pp. 1-32. [ Links ]
Van Vleck, L.D. and Henderson, C.R. (1961). Use of part lactation records in sire evaluation. J. Dairy Sci. 44: 1511-1518. [ Links ]
(Received March 6, 1998)