|Livestock Research for Rural Development 25 (9) 2013||Guide for preparation of papers||LRRD Newsletter||
Citation of this paper
The milk and fat yield data of 28 herds of Holstein cows were collected from 1989 to 2008. First three lactation data of milk and fat yield traits were analyzed by using several univariate and multivariate models. The fixed effect of herd-year-season in all lactations was significant (p<0.01); also, the age at calving and the previous dry period were included in the models as covariate wherever appropriate. Breeding values were calculated by best linear unbiased prediction (BLUP) method with different animal models. Estimation of (co) variance components was performed using WOMBAT software with REML procedure.
Heritability estimates under univariate model for the first, second and third lactation periods were 0.27, 0.19 and 0.13 for milk yield and for the fat yield were 0.18, 0.15 and 0.14, respectively. The heritability values under the repeatability model for milk and fat yield were 0.21 and 0.15, respectively. The repeatability estimates for milk yield was 0.46 and for fat yield was 0.35. Estimated heritabilities using a bivariate model for milk and fat yields were 0.28 and 0.18, respectively; and genetic, phenotypic and environmental correlation between milk and fat yield were 0.84, 0.79 and 0.78, respectively. Heritability estimates for lactations 1 to 3 by multivariate model were 0.28, 0.22 and 0.16 for milk yield, and 0.18, 0.18 and 0.16 for fat yield, respectively. Genetic, phenotypic and environmental correlations between lactation periods were 0.92, 0.55 and 0.43 for first and second lactation; and 0.87, 0.46 and 0.35 for first and third lactation; and 0.97, 0.57 and 0.48 for second and third lactation for milk yield, respectively. These parameters for fat yield were 0.94, 0.43, 0.31 and 0.82, 0.34, 0.23 and 0.93, 0.43, 0.32, respectively.
Keywords: animal model, correlation, heritability, repeatability
Yields of milk, fat, and protein seem to be the main traits considered for dairy cattle selection (Albuquerque et al 1995). A general goal of dairy breeding throughout the world is improved efficiency of milk production (Tekerli et al 2000) that efficient milk production is critical to the success of the dairy industry (Wall and McFadden 2007). Protein and fat yields and somatic cell count (SCC) are important factors determining the farmers return from milk, some of the most commonly used selection indexes, like the USDA Net Merit and Cheese Merit Indexes, are functions of milk constituents and SCC (Castillo-Juarez et al 2002).
Best linear unbiased prediction (BLUP) has become the standard method for inferring breeding values (EBV), especially in dairy cattle. A large number of countries employ an animal model, where an additive genetic effect is fitted for each animal in the pedigree (Urioste et al 2003). To estimate the EBV of animals for selection, variation arise in yield records from systematic environmental effects must be removed. Systematic environmental effects, such as lactation length, age and parity of cow, herd-year-season of calving, duration of the dry period, number of days open and length of the previous and current calving intervals are often included as explanatory variables in the models for genetic evaluation of milk yield (Khan and Shook 1996; Urioste et al 2003).
Knowledge of genetic and phenotypic parameters is required for planning efficient breeding programs in animal husbandry. By knowledge of heritability estimate, animal geneticists can determine whether or not a particular trait can be improved by selection, by improvement of management practices, or both. Selection for yield traits has been successful and has improved the efficiency of dairy production (Dechow et al 2001). Differences in genetic and phenotypic parameter estimates are expected as a result of population structure and estimation methods employed and because of sampling errors (Roman et al 2000). Estimation of co (variance) over time is necessary, because of change in management, increase number of herds and size it, import semen from other countries.
The objectives of this study were to estimate genetic parameters for milk and fat yield traits in Iranian Holstein dairy cattle, and determine efficient model to genetic evaluation of animals.
The data were comprised of production and pedigree databases on Iranian Holsteins of 28 herds in the Khorasan Razavi province that calved from 1989 to 2008. Milk and fat yields of first three lactations were adjusted to 305d and twice daily milking (2×305 days) and only production from complete lactations were analyzed. Age at first calving was defined to be 18 to 40 month of age, the data on heifers that calved out of this range were excluded.
Use of later lactation records would bias the estimates of genetic parameters if a first lactation record was not available (Ojango and Pollott 2001). Hence, an animal was required to have a first lactation record for records from subsequent lactations to be retained for the analysis. This requirement allowed later parity records to be compared with more unselected first parity records and should in part remove some potential bias that may result from dairy farmers being more tolerant of breeding difficulties in higher producing cows (Funk et al 1987).
Second and third lactation records were edited to include cows that had calved no earlier than 10 month after their previous calving and no later than 24 month after their previous calving. Calving intervals less than 10 month may have arisen as a result of abortions or other abnormal occurrences and greater than 24 month were also occur as a result of reproductive abnormality or decline of nutrients. Descriptive statistics for the final data set are presented in Table 1.
Four classes of calving season were defined and herd-year-season contemporary groups (HYS) were formed by combining herd, year and season of calving classes, and only HYS classes with more than five records were retained.
The following animal models were used in order to obtain variance components:
Yijk = µ + HYSi + b1Agej + aj
+ eijk [I]
Yijk = µ + HYSi + b1Agej + b2PDPj + aj + eijk [II]
Yijkl = µ + HYSi + b1Agejk + aj + Pej + eijk [III]
Where, Yijk is the record of observation; µ is the overall population mean; HYSi is the fixed effect of herd-year-season i; b1 is the linear regression coefficient of age at calving of animal j; Agej is the effect of age at calving of animal j (for lactation k) as a covariate; b2 is the linear regression coefficient of previous dry period of animal j; PDPj is the previous dry period of animal j as a covariate; aj is the random additive genetic effect of animal j; Pej is the random permanent environmental effect of animal j and eijk is the random residual effect.
The following analyses were carried out:
Six univariate analyses for each of milk and fat yields in lactations 1-3 using model I for first lactation and model II for second and third lactations; 2 analyses using a repeatability model for each of milk and fat in lactations 1, 2 and 3 using model III; a bivariate analyze for milk and fat yield traits of first lactation using model I; and 2 multivariate analyses for each trait of milk and fat that yield of each lactation consider as separate trait, and model I and II were used for first and later lactations, respectively.
Descriptive statistical analyses were performed using the General Linear Model (GLM) procedure of SAS (SAS Institute 2009). The (co) variance components and genetic parameters of studied traits were estimated with AI-REML algorithm of the WOMBAT software (Meyer 2008).
Variance component and heritability estimates of milk and fat yield traits from univariate animal model analysis are presented in Table 2.
The heritability estimates were slightly lower than those from literature. Most studies suggest that heritability estimates for later lactations cows are lower than estimates for first lactation (Visscher and Thompson 1992; Meyer 1984 ). This decrease is result of increasing the effect of days in milk (DIM) by parity (Barash et al 1996), increasing the effect of days open as herd average increased (Laben et al 1982), increasing in residual variance steadily with lactation number and increasing in other source of environmental variance. Urioste et al (2003) demonstrated that including adjustment for heterogeneous variance and effect of reproductive traits such as dry period and days open in model result in increasing heritability of production traits. Include the days open in models was impossible because insemination date of cows were unavailable; also reported that assumption of heterogeneity of variance is tenable, especially in later lactations.
Heritability difference between first and second lactations is greater than those between second and third lactation. These probably influenced by the common source of variance such as previous dry period and previous days open in later lactations. Heritability estimates will decrease if population was small (Weller et al 1987) and has been under selection several generation (Weller et al 1987; Raheja et al 1989); although any information about selection was unavailable but notice of population mean of yields over years show that the population has been under intensive selection (figures 1 and 2). There were positive trend in two traits over time except for last year because all cows did not complete their lactations and analyzed lactations belong cows with short lactation period.
|Figure 1: Change in milk yield mean over calving year||Figure 2: Change in milk fat yield mean over calving year|
If culling takes place on performance in previous lactations, the parameter estimates from univariate analyses on later lactations will be bias (Visscher and Thompson 1992). In this study 36% of animals in first lactation don’t have any record in second lactation, and 42% of animals in second lactation don’t have any record in third lactation, only 37% of animals have records in three lactations.
Variance components, heritability and repeatability for milk and fat yield are given in Table 3. Heritability of milk yield is within the range of estimates from literature but heritability of fat yield is lower. Coefficient of variation (CV) for milk and fat yields were 22% and 23%; respectively. These values indicate that phenotypic variances are high and caused lower heritability. High coefficient of variation in production is due to intensive change in weather of Khorasan Razavi province such as warm and dry summer and cold winter. Due to high ratio of the permanent environmental to the phenotypic variance (c²), estimated repeatability for milk and fat yields were high and are in the range of estimates from literature. Because repeatability is correlation between repeated records, selection on first lactation result to increase production mean in subsequent lactations. High difference between heritability and repeatability (r - h²=c²) indicated that non-additive genetic and permanent environment had more influence than additive genetic on these traits.
High ratio of the residual to the phenotypic variance (σe² / σp² =1-r) in these traits indicated that except for all effects in model, there are unknown effects such as recording system and its accuracy, difference in feeding system, etc that had high effect on production. In general, in Iran does not use advanced methods to selection, and selection criteria is phenotypically, therefore it is possible to remain cows in herds in later lactations with high production because of their gene combination value not breeding value. This can decrease heritability over years.
Dong et al (1988) investigated the effect of the degree of completeness of the relationships in models which considered relationships among animals and found that heritability estimates from REML were lower if the relationships were from sires only, compared with those from more complete pedigrees. They also found that full relationships with REML from ancestors of about two generations resulted in slightly higher estimates than when relationships were from only one generation. Also Urioste et al (2003) found low heritability influenced by the weak structure of the pedigree information. Any individual without records connected to only one another animal in the pedigree does not add any information (Meyer 2003). In this study about 52% of animals were without records connected to only one another animal, and 53% of animals were grade.
Co (variance) components, genetic, residual and phenotypic correlations between milk and fat yields of first lactation obtained from bivariate analysis are in Table 4. Variance components and heritability of both traits were slightly greater than those estimated from univariate analyses. Additive genetic variances about 2.6% and <1% were greater than those from univariate analysis for milk and fat yields; respectively.
When the heritability, genetic and environmental correlations for two traits are equal, multivariate predictions are equivalent essentially to those from univariate analysis for each trait. Moreover, traits with lower heritability benefit more when analyzed with traits with higher heritability in a multivariate analysis. Also, there is an additional increase in accuracy with multivariate analysis resulting from better connections in the data due to residual covariance between traits (Mrode 2005).
Albeit heritability of both traits from univariate analyses were different, but different between genetic and environmental correlations (0.063) were small that resulting in slightly increase in heritability. It was expected to increase heritability of milk fat yield more than it for milk yield because of its low value, but increase in heritability of milk yield was greater. This is due to small different of genetic and environmental correlations. Estimated correlations are in the range of estimates from literature.
There was a very high positive genetic correlation of 0.848 between milk and fat yields suggesting that the same set of genes exerts a common influence on the both traits and selection for milk yield would increase fat yield. High positive environmental correlation of 0.785 indicated that both traits were affected by same environmental effects.
Co (variance) components, heritability and correlations between yields of lactations are shown in Tables 5 for milk yield and 6 for fat yield. Correlations of both traits and heritability for milk yields were in the range of estimates from literature but heritability for fat yields were lower. Decreasing in heritability of production traits across lactations was observed (Visscher and Thompson 1992).
Comparison of estimated variance components from univariate and multivariate analysis were showed that there are no increasing (almost 1.9%) in additive genetic variances in first lactation for both traits but these increase were about 24.1% and 42.4% for milk yield of second and third lactations and about 19.8% and 26.4% for fat yield of second and third lactations. This result means that first lactation records are without culling bias and second and third lactation records are influenced by culling bias.
Except for heritability of milk fat of first lactation from univariae analysis that equal to those estimated from multivariate analysis, other estimated heritability from multivariate analyses are larger than those estimated from univariate analyses. Using multivariate analysis, the heritabilities were higher than those estimated by the univariate analyses, possibly due to correction of selection bias made by multivariate analysis. High producing heifers in low management herds are unable to recover properly after first lactation, and therefore, do not realize their full potential during second lactation, thus reducing the genetic variance of multiparous cows (Weller et al 1987).
Estimated phenotypic, additive genetic and residual correlations were showed that correlations between consecutive lactations (first and second; second and third) were larger than those between nonconsecutive lactations (first and third). This outcome is in agreement with the findings of Visscher and Thompson (1992).
In this research, the differences in genetic and residual correlations among lactations varied between 0.49 and 0.63, so application of multivariate analysis is valid according to Schaeffer (1999), who suggested the feasibility of this type of analysis for characteristics where there are large differences between the genetic and residual correlations, preferably greater than 0.50.
Different result from several analyses and their effects on variance components were presented that use of these analyses without attention to their individual characterization will result in unexpected outcome in genetic trend. Genetic analysis by univariate model are easy but its result is only unbiased for the first lactation because only result of first lactation is in agreement with those in multivariate analyses. Selection on estimated breeding value of later lactations from univariate analyses will be biased (culling bias) because doesn’t consider prior selection criteria.
One of the assumptions behind the use of repeatability models is that all lactations are genetically the same trait. This assumption is supported by observations that the genetic correlations among records are close to unity (Tong et al 1979; Meyer 1984 1985). However, some authors have claimed that lactations should be considered as separate traits because genetically they may be different entities. For this reason, they suggest the use of multivariate analysis (Alburqueque et al 1996).
Difference between genetic correlations and difference in variance components from multivariate analyses of each trait for lactation yields were showed that different lactations are separate traits and aren’t in influence of the same set of genes.
Heritability values of both traits from repeatability model were lower than those estimated from univariate analyses (average of first three lactations). This might be explained because of individual characterization of repeatability model, important sources of variance like open days, dry period and other factors that affect only on later lactations can’t include in model.
Butcher and Freeman (1968), reported that the repeatability of consecutive lactations increased gradually as the animals got older whereas that of nonconsecutive lactations decreased gradually as the lactations became more separated in time, also De Varies et al (1998) demonstrated that use of repeatability model to genetic evaluation cause to select older animals. These cause to increasing generation interval and decreasing in genetic gain.
Based on the results of this research it is concluded that:
Albuquerque L G, Dimov G and Keown J F 1995 Estimates using an animal model of (Co) variances for yields of milk, fat, and protein for the first lactation of Holstein cows in California and New York. Journal of Dairy Science 78: 1591-1596. http://dx.doi.org/10.3168/jds.S0022-0302(95)76783-2
Albuquerque L G, Keown J F and Van Vleck L D 1996 Genetic parameters of milk, fat and protein yields in the first three lactations, using an animal model and restricted maximum likelihood. Brazilian Journal of Genetics 19: 79-86.
Barash H, Silanikove N and Weller J 1996 Effect of season of birth on milk, fat, and protein production of Israeli Holsteins. Journal of Dairy Science 79: 1016–1020.
Butcher D F and Freeman A E 1968 Heritabilities and repeatabilities of milk and milk fat production by lactations. Journal of Dairy Science 51: 1387-1391.
Castillo-Juarez H, Oltenacu P A and Cienfuegos-Rivasc E G 2002 Genetic and phenotypic relationships among milk production and composition traits in primiparous Holstein cows in two different herd environments. Livestock Production Science 78: 223–231.
Dechow C D, Rogers G W and Clay J S 2001 Heritabilities and correlations among body condition scores, production traits, and reproductive performance. Journal of Dairy Science 84: 266–275.
De Varies M J, Van Der Waaij E H and Van Arendowk J A M 1998 Estimation of genetic parameters for litter size in sheep: A comparison of repeatability and a multivariate model. Animal Science. 66: 685-688.
Dong M C, Van Vleck L D and Wiggans G R 1988 Effect of relationships on estimation of variance components with an animal model and restricted maximum likelihood. Journal of Dairy Science 71: 3047-3052.
Funk D A, Freeman A E and Berger P J 1987 Effects of previous days open, previous days dry, and present days open on lactation yield. Journal of Dairy Science 70: 2366-2373.
Khan M S and Shook G E 1996 Effects of age on milk yield: Time trends and method of adjustment. Journal of Dairy Science 79: 1057-1064.
Laben R C, Shanks R D, Berger P J and Preemaa A E 1982 Factors affecting milk yield and reproductive performance. Journal of Dairy Science 65: 1005-1015.
Meyer K 1984 Estimates of genetic parameters for milk and fat yield for the first three lactations in British Friesian cows. Animal Production 38: 313-332.
Meyer K 1985 Genetic parameters for dairy production of Australian black and white cows. Livestock Production Science 12: 205-219.
Meyer K 2003 Secateurs to prune your pedigrees. Animal Genetics and Breeding Unit, University of New England, Armidale, NSW 2351, Australia.
Meyer K 2008 WOMBAT A program for Mixed Model Analyses by Restricted Maximum Likelihood. USER NOTES. Version 1.0, Animal Genetics and Breeding Unit, University of New England, Armidale, NSW 2351, Australia.
Mrode R A 2005 Linear models for the prediction of animal breeding values (2nd ed). CAB International, Wallingford, UK.
Ojango J M K and Pollott G E 2001 Genetics of milk yield and fertility traits in Holstein-Friesian cattle on large-scale Kenyan farms. Journal Animal Science 79: 1742–1750.
Raheja K L, Burnside E B and Schaeffer L R 1989 Relationships between fertility and production in Holstein dairy cattle in different lactations. Journal of Dairy Science 72: 2670-2678.
SAS 2009 User’s
guide, Release 9.2. SAS Institute Inc., Cary, NC, USA.
Schaeffer L R 1999 Multiple traits animal models. Univ. Guelph. Disponible en: http://www.aps.uoguelph.ca/~lrs/Animalz/lesson15 (August 24, 2005).
Tekerli M, Akinci Z, Dogan I and Akcan A 2000 Factors affecting the shape of lactation curves of Holstein cows from the Balikesir province of Turkey. Journal of Dairy Science 83: 1381–1386.
Tong A K W, Kennedy B W and Moxley J E 1979 Heritabilities and genetic correlations for the first three lactations from records subject to culling. Journal of Dairy Science 62: 1784-1790.
Urioste J I, Rekaya R, Gianola D, Fikse W F and Weigel K A 2003 Model comparison for genetic evaluation of milk yield in Uruguayan Holsteins. Livestock Production Science 84: 63–73.
Visscher P M and Thompson R 1992 Univariate and multivariate parameter estimates for milk production traits using an animal model. I. Description and result of REML analyses. Genetic Selection Evolution 24: 415-430.
Wall E H and McFadden T B 2007 The milk yield response to frequent milking in early lactation of dairy cows is locally regulated. Journal of Dairy Science 90: 716–720.
Weller J I, Ron M and Bar-Anan R 1987 Effects of persistency and production on the genetic parameters of milk and fat yield in Israeli-Holsteins. Journal of Dairy Science 70: 672-680.
Received 8 May 2013; Accepted 3 Aug 2013; Published 4 September 2013
Go to top