| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
,2,3
* Cooperative Research Centre for Cattle and Beef Quality, CSIRO Livestock Industries, North Rockhampton QLD 4702, Australia and
and
Animal Genetics and Breeding Unit,4, University of New England, Armidale NSW 2351, Australia
3 Correspondence:
E-mail:
djohnsto{at}metz.une.edu.au.
 |
Abstract |
|---|
 Top Abstract IntroductionMaterials and MethodsResults and DiscussionImplicationsLiterature Cited |
|---|
Key Words: Beef Cattle • Crossbreeding • Genetic Parameters • Genetic Variance
|
Introduction |
|---|
|
TopAbstractIntroductionMaterials and MethodsResults and DiscussionImplicationsLiterature Cited |
|---|
Procedures for estimating across-breed EBV in the form of across-breed adjustment tables were first presented by Notter and Cundiff (1991). These procedures, while ad hoc in nature, highlighted the need for well-designed crossbreeding studies to augment field data (Rodríguez-Almeida et al., 1997). Alternatively, multibreed evaluations (MBE) are a logical extension of adjustment tables to more fully characterize breeds and breed crosses. A number of models have been proposed for MBE (e.g., Elzo and Famula, 1985; Arnold et al., 1992) and reviewed by Pollak and Quaas (1998).
The present study describes an experimental design that can contribute data for the development of MBE in Australia. The objectives of this study were to analyze growth and carcass data from purebred and crossbred progeny to estimate 1) sire breed means, 2) components of across-breed adjustment factors, and 3) genetic correlations between purebred and crossbred performance.
|
Materials and Methods |
|---|
|
TopAbstractIntroductionMaterials and MethodsResults and DiscussionImplicationsLiterature Cited |
|---|
The CRC crossbred project was based on approximately 1,000 Brahman females mated to sires of nine breeds representing an array of performance and adaptation potential. First mating occurred in 1995 and the next 3 yr in two herds in central Queensland. The crossbred data (XBRED, n = 1,241) used in this study were calves from six sire breeds (AA, HH, SS, BB, BR, and SG) that were also used in the PBRED project. A total of 916 first-crosses and 325 Brahman controls, including both steers and heifers, were produced. Approximately 40% of XBRED cattle of all sire breeds were finished after weaning in the temperate northern New South Wales locations, with the remainder finished in subtropical environments of central Queensland.
For both PBRED and XBRED projects, cattle were managed under two finishing regimens (pasture and feedlot) to finish at representative market live weights of 400 (domestic), 520 (Korean), and 600 kg (Japanese). Slaughter occurred when the average of the cohort (year, season, finish, location, and market) reached the target market weight.
Sires that produced progeny in both the PBRED and XBRED projects are referred to as connecting sires. Table 1
displays the number of sires and progeny by breed and project. For example, in the PBRED project, 115 Angus sires produced 1,829 progeny. Of those 115 sires, 9 where used in the XBRED project, producing 137 crossbred (AA x BB) progeny and 342 straightbred progeny in the PBRED project.
|
) were born between 1984 and 1993 and all had across-herd, within-breed group BREEDPLAN EBV. Data collected from XBRED were not used to generate within-breed EBV. The 1999 group BREEDPLAN evaluation of each breed was used to obtain the sires’ EBV for 400-d weight, carcass weight, retail beef yield percentage, intramuscular fat percentage, rump fat depth, and carcass eye muscle area. All genetic evaluations were performed using the analytical model of BREEDPLAN Version 4.1 (Johnston et al., 1999). Carcass trait EBV were mostly predicted through genetic relationships between live animal ultrasound scans of yearling seedstock bulls and heifers and abattoir measurements from finished steers and heifers (Reverter et al., 2000).
Description of Data
Live weight measurements were taken at around 400 d of age (400W) at the end of the backgrounding period. Protocols and descriptions of all carcass measures are presented in Perry et al. (2001). Traits included hot carcass weight (CWT), retail beef yield percentage (RBY), intramuscular fat percentage (IMF), subcutaneous rump fat depth at the P8 site (P8), and preslaughter ultrasound scanned eye muscle area (SEMA). In brief, IMF was measured with near-infrared spectroscopy using a Technicon Infralyser 450 (Bran+Luebbe, Homebush, NSW, Australia). Within a region, SEMA was measured by the same technician using an Aloka 500V scanning machine using a 3.5-MHz, 17.2-cm linear array transducer (Corometrics Medical System, Wallingford, CT). Although differences between scanning technicians were accounted for in all analyses by the fixed contemporary group effect, any technician bias in measuring SEMA will still be present in raw means. Measurements of RBY were obtained as the total weight of 17 trimmed boneless retail primal cuts, plus the weight of adjusted manufacturing trim, expressed as a percentage of the recovered left side weight. Measurements on RBY, IMF, and P8 were taken from the left side of the carcass only. Table 2 shows the number of observations, unadjusted means, and SD for all traits.
|
A preliminary analysis was performed to ensure that subsequent estimates of genetic parameters would not be biased due to nonrandom allocation of sire breed across CG. To this end, the mean sire breed EBV, expressed as deviations from their respective CG means, were compared. The model was the same as the one used to compute LSM, except that the random additive genetic effect of animal was included. Also, Brahman calves were removed from this analysis because they did not contribute to the calculation of purebred-crossbred correlations. Ancestors were traced back five generations to generate a pedigree of 2,281 animals. There was no CG with individuals from a single sire breed. The percentage of CG with individuals from two, three, four, and five sire breeds was 4.8, 7.9, 25.4, and 61.9, respectively. Analyses were performed using VCE 4.2.5 (Groeneveld and García-Cortés, 1998). The multiple range test devised by Duncan (1955) and extended by Kramer (1957) was used to test for significant breed differences among three or more means.
Across-Breed Adjustments
Components required to estimate across-breed adjustment factors were developed for all traits. Following Cundiff (1994), data for trait i on crossbred progeny were adjusted to the 1997 birth year for sire breed j by the following:
|
|
where
is the LSM for trait i and sire breed j adjusted to 1997, ßi is the pooled within-breed regression coefficient of crossbred calf performance on sire EBV for trait i, EBV(np)ij is the 1999 nonparent mean EBV for trait i and breed j (i.e., the average 1999 EBV for calves born in 1997), and EBV(s)ij is the mean 1999 EBV of sires weighted by the number of crossbred progeny each sire produced.
Regression coefficients for each trait (ßi) were obtained using GLM in SAS. Individual crossbred performance for each trait was the dependent variable in a model that included sire EBV for that trait and age at measurement as covariates together with breed and CG as fixed effects. Each regression coefficient is expected to have a value of 0.5 for each trait. Deviations from 0.5 were used to adjust sire breed differences for differential effects of scaling on performance of crossbred progeny.
Choosing Angus as the base breed, across-breed adjustment factors for trait i and breed j (Aij) can be calculated as (Cundiff, 1994):
Genetic Correlation Between Purebred and Crossbred Performance (rpc)
Purebred-crossbred genetic correlations were estimated for each trait to explore the relative emphasis to give purebred and crossbred information when selecting for crossbred performance as well as to explore possible re-ranking of sire’s performance when used in crossbreeding programs. Purebred and crossbred performance were considered as different traits. Multivariate analyses were applied to estimate their heritabilities (h2) and the genetic correlation between them. For each trait, the vectors yP and yx contain observations in PBRED and XBRED, respectively. The bivariate animal model can be represented as follows:
 |
with expectation
 |
and variance
|
|
where XP(XX) is a known incidence matrix relating observations in yP(yX) to age covariate and CG fixed effects in vector ßP(ßX); ZP(ZX) is a known incidence matrix relating observations in yP(yX) to random additive genetic values in uP(uX); eP(eX) are unknown vectors of random temporary environmental effects; A is Wright’s numerator relationship matrix between all animals; I is an identity matrix;
is the additive genetic variance for purebred (crossbred) performance;
is the additive genetic covariance between purebred and crossbred performance; and
is the residual variance for purebred (crossbred).
Definition of CG to estimate heritabilities and correlations were as defined earlier but also included sire breed. The number of CG for 400W, CWT, RBY, IMF, P8, and SEMA were 432, 1,192, 521, 1,136, 1,125, and 1,057, respectively, for PBRED and 96, 331, 95, 320, 325, and 311, respectively, for XBRED. Analysis of PBRED data included observations from progeny of all sires (Table 1) except Brahman calves. Ancestors were traced back five generations to generate a pedigree of 17,789 animals. Genetic parameters were estimated by REML using analytical gradients with VCE 4.2.5 (Groeneveld and García-Cortés, 1998).
|
Results and Discussion |
|---|
|
TopAbstractIntroductionMaterials and MethodsResults and DiscussionImplicationsLiterature Cited |
|---|
because data sets are confounded with herds of origin and backgrounding and finishing locations, and to some extent sex. For example, in the PBRED data set no AA, HH, or SS calves were produced in the subtropical northern locations. In addition, heifers accounted for 46% of XBRED animals but only 27% in the PBRED. The high SD (74.1 d) observed for 400W age in the PBRED data set was attributed to animals being purchased at weaning from several cooperating seedstock herds. In contrast, XBRED individuals were born at CRC locations, allowing for a more homogeneous age distributions, yielding smaller SD (46.1 d).
Mean sire EBV (expressed as deviations from CG) were estimated from the XBRED dataset to assess the randomness of sire allocations across CG. The results showed no significant differences among sire breeds’ EBV deviations. This suggested that sires were effectively randomly allocated across CG. The only exception was that the mean EBV deviation for IMF of crossbred calves out of SS sires was higher than that of SG-sired calves. Also, when sire breed was excluded from the model the resultant mean EBV deviations were equivalent to crossbred performance comparisons from Table 3 (data not included). Levene’s test (Milliken and Johnson, 1984) rejected (P < 0.1) the hypothesis of homogeneity of variances in EBV deviations for IMF, P8, and SEMA, anticipating heteroskedasticity at the additive genetic level across the different sire breeds for these traits. A similar approach to compare parental breeds in crossbred sheep was reported by Sakul et al. (1999), who failed to observe significant differences in mean EBV among crossbred ewes for total milk yield and percentage lactose.
|
presents LSM for each trait by sire breed for the XBRED data sets. For all traits, heterosis could not be estimated from the data and therefore sire breed means may contain nonadditive effects. Crossbred progeny of AA and HH sires had the heaviest CWT and largest SEMA. In contrast, progeny of BR and BB sires had the lightest carcasses. Angus-cross progeny had the highest P8 and IMF with lower RBY, whereas BB and SG-cross progeny had the lowest IMF. These results are in agreement with Newman et al. (1999a, b) from a study with a subset of these data.
Regressions of crossbred calf performance on sire EBV were all positive and significantly different from zero. The regression of 400W on 400W EBV was 0.588 ± 0.190 kg/kg and did not differ significantly from the expected 0.5 kg/kg. Similar results were found for CWT (0.492 ± 0.101 kg/kg), RBY (0.485 ± 0.164 %/%), and SEMA (0.462 ± 0.112 cm2/cm2). Van Vleck and Cundiff (2000) reported coefficients of 0.57, 0.57, and 0.54 (lb/lb) when regressing F1 yearling weight on sire EBV in Hereford, Angus, and Shorthorn, respectively. In contrast, regressions of crossbred IMF on sire IMF EBV and crossbred P8 on sire P8 EBV were 0.263 ± 0.079 %/% and 0.219 ± 0.141 mm/mm, respectively. These values differed from 0.5 (P < 0.10), suggesting that sire genetic differences in IMF and P8 were not being fully expressed in the crossbred progeny. This may also be due to the EBV’s being computed to predict differences at a standard 300-kg carcass weight basis. The mean CWT of crossbreds in this study was only 259 kg (Table 2), and therefore, on average, the regressions of crossbred carcass performance on carcass EBV would be underestimated.
Adjustment factors to allow comparison of animals across breeds (Notter and Cundiff, 1991) were computed to adjust for sampling of sires and genetic trend. Table 4 presents the mean EBV for each trait for the sires used in XBRED project, weighted by the number of crossbred progeny, from each breed’s 1999 BREEDPLAN genetic evaluation. Mean EBV for all 1997-born calves (i.e., nonparents) from the same evaluation (1999) are also presented. With the exception of SS, the average EBV of the sires used in the XBRED project were above breed average for 400W and CWT. Further, HH sires used were below breed average for carcass fat traits (IMF and P8).
|
) and the 1997 average EBV for each breed (Table 4) using the appropriate pooled within-breed regression coefficients and equations reported above. These adjustments allowed sire breed means to be compared on a 1997 basis (Table 5). The year 1997 was chosen as the base from 1999 group BREEDPLAN genetic evaluations for each breed simply as a point in time comparison for this work. It is also the method used for expressing breed average EBV for each of the breeds in their industry publications. Considerable sire breed differences existed after adjusting for the sample of sires. However, still contained in these breed differences are any hybrid vigor effects. It would be expected that the British breeds would express more hybrid vigor in their progeny from Brahman cows compared to the SG and BR breeds. Van Vleck and Cundiff (2000) reported Hereford, Angus, and Shorthorn breed means, adjusted for sire sampling, that only differed by 6 kg for yearling weight across the three breeds. These results are very similar to the magnitude of the differences estimated in our study for those same breeds. However, no breed estimates were presented for carcass traits or for SG or BR breeds in their work.
|
presents estimates of additive genetic variance, heritability (h2), and purebred-crossbred genetic correlations (rpc). Crossbreds expressed greater genetic variation and higher h2 for 400W and CWT compared to the purebreds. The comprehensive review of h2 estimates in beef cattle published from 1946 to 1991 of Koots et al. (1994) reported a weighted mean h2 estimate of 0.33 for yearling weight and 0.23 for CWT on an age-constant basis and 0.48, 0.36, 0.46, and 0.41 for cutability, marbling, backfat, and eye muscle area on a weight-constant basis, respectively. These h2 are similar to those estimated in our study for the related traits of RBY, IMF, P8, and SEMA, in both PBRED and XBRED. The h2 estimates for 400W and CWT in both PBRED and XBRED are higher than those reported in Koots et al. (1994).
|
The less than unity correlation observed for some of the traits may simply be the result of low numbers, particularly for the crossbreds. Also, the datasets used to estimate the genetic correlations were much smaller than the data used by the BREEDPLAN analyses to compute the EBV used in the regression analyses. However, one possible explanation for the less than unity correlation is given by Wei and van der Werf (1995). They reported, in general, rpc decreases with increasing dominance level or gene frequency difference between parental populations, whereas a high positive rpc is often indicative of the greater importance of additive genetic variance. It could be possible that the traits that expressed the rpc less than unity (400W, CWT, and SEMA) are traits that have been selected for several years (especially CWT and SEMA as correlated responses to selection for growth), whereas RBY, IMF, and P8 expressed higher rpc and are likely to have had a shorter selection history in most of the breeds due to the relatively recent inclusion of carcass traits (particularly RBY and IMF) in Australian genetic evaluations (Johnston et al., 1999).
Another factor that may have contributed to the less than unity correlation is that the crossbreds were reared preweaning (and some animals for whole of life) in a subtropical environment. This environment is likely to be quite different from the average environment of the performance data used in the Angus, Hereford, and Shorthorn BREEDPLAN and may reflect a genotype x environment interaction. Also, all of the crossbreds were half-Brahmans and had the preweaning influence of the Brahman female that may have also contributed to differences in the expression of these traits.
For traits such as 400W, CWT, and SEMA, for which the estimates of the rpc were less than unity, consideration will need to be made when using these data for genetic improvement. In addition, for traits exhibiting heterosis, consideration of this effect will be required, particularly when purebreds and crossbreds are reared together. Armstrong et al. (1994) concluded that the inclusion of crossbred data significantly increased the accuracy of evaluations. Traits that exhibit low rpc could be used in a genetic evaluation, wherein crossbred data could be treated as different traits and EBV reported within-breed (purebred level) only, which differs from the approach of Armstrong et al. (1994).
Although our work allowed the calculation of across-breed adjustment factors using the method of Cundiff (1994), we chose not to present these given our previous discussion regarding the low regression for carcass traits and the less than unity correlations for growth traits. Caution must be taken in the generation of across-breed EBV calculated by adjustment factors. The across-breed EBV are less accurate for comparing bulls of different breeds than the comparison of bulls within-breed. The accuracy of across-breed EBV depends on the accuracy of the within-breed EBV and, to a lesser extent, the accuracy caused by error in the estimation of adjustment factors. This is of particular concern when dealing with carcass EBV. To date, the majority of data used in the calculation of carcass EBV has come from ultrasound scan measurements taken by accredited scanning technicians. Without abattoir measurements, an upper bound on the accuracy of any carcass EBV will be given by the genetic correlation between live-animal ultrasound scans and abattoir measurements. For instance, the 1999 spring Angus group BREEDPLAN analysis included 167,961 measurements on 400-d weight and 57,924 live scans of P8 fat depth but only 1,497 carcass P8 fat from the abattoir. Recent estimates of genetic correlation between live-animal ultrasound scans and abattoir measurements are favorable and moderately strong (Moser et al., 1998; Reverter et al., 2000).
Many seedstock breeders in Australia register animals and performance information on cattle with more than one breed society. However, these data are not recorded in such a way that allows the combining of different breed society databases (Graser, 1999). Unless previously specified, the assumption is made that animals from different breeds are run in separate management groups. This does not allow for direct comparison of performance. Often, management group effects are confounded with breed and heterosis, resulting in F1 progeny not being properly evaluated (Klei et al., 1996). Therefore, there is a need for performance data from crossbreeding experiments that have direct breed comparisons, to be used to augment field data to allow the generation of multibreed genetic evaluations.
|
Implications |
|---|
|
TopAbstractIntroductionMaterials and MethodsResults and DiscussionImplicationsLiterature Cited |
|---|
|
Footnotes |
|---|
2 Present address: PIC-USA, PO Box 348, Franklin, KY 42135.
4 AGBU is a joint institute of NSW Agriculture and The Univ. of New England.
Received for publication July 9, 2001. Accepted for publication February 1, 2002.
|
Literature Cited |
|---|
|
TopAbstractIntroductionMaterials and MethodsResults and DiscussionImplicationsLiterature Cited |
|---|
This article has been cited by other articles:
|
|
 |
|
  D. G. Riley, S. W. Coleman, C. C. Chase Jr., T. A. Olson, and A. C. Hammond Genetic parameters for body weight, hip height, and the ratio of weight to hip height from random regression analyses of Brahman feedlot cattle J Anim Sci, January 1, 2007; 85(1): 42 - 52. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
), heritabilities (h2) and purebred-crossbred genetic correlations (rpc) estimated from combined purebred and crossbred data