| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
ANIMAL GENETICS |
* Agriculture and Agri-Food Canada Research Centre, Lethbridge, Alberta T1J 4B1, Canada and
and
Department of Animal Sciences, Cornell University, Ithaca, NY 14853
Abstract
This study was conducted to compare carcass EPD predicted using yearling live animal data and/or progeny carcass data, and to quantify the association between the carcass phenotype of progeny and the sire EPD. The live data model (L) included scan weight, ultrasound fat thickness, longissimus muscle area, and percentage of intramuscular fat from yearling (369 d of age) Simmental bulls and heifers. The carcass data model (C) included hot carcass weight, fat thickness, longissimus muscle area, and marbling score from Simmental-sired steers and cull heifers (453 d of age). The combined data model (F) included live animal and carcass data as separate but correlated traits. All data and pedigree information on 39,566 animals were obtained from the American Simmental Association, and all EPD were predicted using animal model procedures. The genetic model included fixed effects of contemporary group and a linear covariate for age at measurement, and a random animal genetic effect. The EPD from L had smaller variance and range than those from either C or F. Further, EPD from F had highest average accuracy. Correlations indicated that evaluations from C and F were most similar, and L would significantly (P < 0.05) re-rank sires compared with models including carcass data. Progeny (n = 824) with carcass data collected subsequent to evaluation were used to quantify the association between progeny phenotype and sire EPD using a model including contemporary group, and linear regressions for age at slaughter and the appropriate sire EPD. The regression coefficient was generally improved for sire EPD from L when genetic regression was used to scale EPD to the appropriate carcass trait basis. The EPD from C and F had similar linear associations with progeny phenotype, although EPD from F may be considered optimal because of increased accuracy. These data suggest that carcass EPD based on a combination of live and carcass data predict differences in progeny phenotype at or near theoretical expectation.
Key Words: Beef Cattle • Carcass • Genetic Evaluation • Simmental • Ultrasound
Introduction
Several breed associations now conduct national cattle evaluations for carcass merit using live weights and real-time ultrasound (RTU) measures of body composition from yearling bulls and heifers to augment carcass data collected on progeny of sires. The inclusion of live animal data enables the evaluation of a larger, more random sample of the population and increases the accuracy with which potential replacements are evaluated. Crews et al. (2003)
reported genetic parameter estimates among live animal measurements of yearling replacements and carcass traits of market progeny in the Simmental breed.
In a multivariate carcass merit evaluation, carcass and live animal measures are considered separate but correlated data. Through additive relationship and genetic correlation, such an evaluation results in EPD for both live and carcass traits. Crews and Kemp (2002) noted that the addition of live animal to carcass data resulted in the most significant accuracy increases for EPD of young replacements that had not yet produced progeny with carcass data.
Crews (2002) noted that studies of the relationship between sire carcass EPD and progeny phenotype were lacking but showed that differences in sire carcass EPD were related to differences in progeny phenotypes at or near theoretical expectation in Charolais. This relationship has not been investigated with respect to carcass EPD based on combinations of live and/or carcass data. Genetic (co)variance estimates in the report by Crews et al. (2003) and other recent studies (Reverter et al., 2000; Crews and Kemp, 2001; Devitt and Wilton, 2001), indicated that mean and variance differences between carcass traits and their live animal indicators suggest that EPD based on combinations of carcass and/or live animal data should be compared. The objective of this study was to compare carcass EPD estimated using live and carcass data alone and in combination, and to estimate their association with differences in progeny phenotypes.
Materials and Methods
Breeding Value Estimation.
Three sets of EPD were estimated for 39,566 animals with ties to live animal and/or carcass data in the American Simmental Association (ASA) performance database. Traits of interest included hot carcass weight (HCW), subcutaneous fat thickness (FAT), longissimus muscle area (REA), and marbling score (MAR). Model 1 included only live (L) animal data from yearling bulls and heifers and EPD were estimated for scan weight (SWT), RTU fat thickness (RFAT), RTU longissimus muscle area (RREA), and RTU percentage of intramuscular fat (RIMF). Model 2 included only carcass (C) data from steers and cull heifers, and EPD were estimated for HCW, FAT, REA, and MAR. Model 3 (F) included both live and carcass data, and similar to C, EPD for HCW, FAT, REA, and MAR were estimated with F. With Models L and F, the RTU data from yearling bulls and heifers were considered separate, genetically correlated traits, whereas the SWT of bulls and heifers were considered genetically equivalent (Crews and Kemp, 2001; Crews et al., 2003). All indicator traits measured on live animals were considered separate from but correlated with carcass traits (Reverter et al., 2000; Crews and Kemp, 2001; Crews et al., 2003). Variance components and genetic parameters were estimated by Crews et al. (2003) and are summarized in Table 1. All EPD and accuracy values from L, C, and F were estimated using software tools in the Animal Breeder’s Tool Kit (B. L. Golden, personal communication). Although EPD are computed for RTU traits on both bull and heifer bases, only bull-basis EPD will be presented; selection would be expected to be primarily of bulls, and the report of Crews et al. (2003) showed that genetic correlations of bull RTU traits with corresponding carcass traits tended to be equal to or higher than those involving heifer RTU traits.
|
), reference animals were the 202 sires with at least 15 progeny with either live animal or carcass data. Evaluations resulting from L, C, and F were compared among the reference animals using mean and correlation procedures of SAS (SAS Inst., Inc., Cary, NC).
EPD Validation.
Carcass EPD predict genetic differences among sires for component traits, such as carcass weight, longissimus muscle area, subcutaneous fat thickness, and marbling score. Few studies have reported the association between sire EPD and progeny phenotype for carcass traits (Crews, 2002). Carcass records from 824 steers and heifers sired by 84 bulls with EPD from L, C, and F were subjected to general linear models procedures of SAS. The base model included the effect of slaughter contemporary group (slaughter date x sex x percentage of Simmental [50, 62, 75, 94%]) and the linear regression of age at slaughter (465 d, SD = 62 d). Models for HCW, FAT, REA, and MAR included the sire carcass EPD for the corresponding traits as a linear regression. Three models were fit for each trait in order to compare the association between sire carcass EPD from L, C, and F with progeny phenotype. Carcass records on these progeny were added to the ASA database in the year following those included in the report of Crews et al. (2003); therefore, all EPD were computed without these data included. The hypothesis that the regression of progeny phenotype on sire EPD was different from 0 as well as equal to the theoretical expectation of 1 was tested for each of the four traits and three sets of sire EPD using methods described by Basarab et al. (1994) and used by Crews (2002).
A modification of this procedure was required for sire EPD from L. Because L did not include carcass data, sire EPD were estimates based entirely on live animal data, leading to an expectation for the sire EPD regression that is not equal to 1. However, it can be shown that, in a bivariate model, when data are available on only one trait, EPD for the second trait are equal to predictions from genetic regression (Cameron, 1997), represented as
 |
where û2, the EPD for the second, unmeasured trait, is equal to the EPD for the first, measured trait (û2) multiplied by the genetic regression coefficient, defined as the ratio of the genetic covariance (
g1,g2) to the genetic variance of the measured trait (
). Genetic (co)variances required for this prediction were taken from Crews et al. (2003) (Table 1
).
In order to unambiguously refer to EPD from the three models, as well as EPD predicted by genetic regression, EPD and models in this study will be designated explicitly by both trait and model. For example, HCW EPD from F are denoted HCW-F, and SWT EPD from L denoted SWT-L. The EPD denoted HCW-R, FAT-R, REA-R, and MAR-R refer to hot carcass weight, fat thickness, longissimus muscle area, and marbling score EPD predicted by genetic regression from SWT-L, RFAT-L, RREA-L, and RIMF-L, respectively.
Results and Discussion
Summary Statistics for EPD and Accuracy.
Live and carcass EPD from L, C, and F are summarized for the 202 reference sires in Table 2. The EPD from L, based entirely on live animal data, had smaller range than EPD from the model based only on carcass data or the combined data model. The ranges of EPD from L were 73, 42, 77, and 49% of those for the corresponding EPD from F for weight, fat thickness, longissimus muscle area, and percentage of intramuscular fat. Further, the ranges of EPD from L were from 48 to 75% of those from C. It was expected that EPD from L would have different first and second central moments from those from C and F due to differences in age at measurement, gender, and variance components. Live animal measurements were taken on yearling bulls and heifers, whereas carcass traits were measured on steers and cull heifers that were older (453 d of age) than yearling. Additionally, yearling RTU percentage of intramuscular fat and carcass marbling score are measured using different units with inherently different mean and variance apart from differences in age and gender. The smaller range and SD of EPD L suggest that yearling live animal measures, especially among bulls, exhibited less variability than EPD from models that included carcass data. Between C and L, EPD were based on mutually exclusive data, and heritability estimates for yearling live animal measurements (0.47, 0.53, 0.37, and 0.47 for SWT, RFAT, RREA, and RIMF, respectively) were similar to or higher than those for carcass traits (0.48, 0.35, 0.46, and 0.54 for HCW, FAT, REA, and MAR, respectively). Therefore, differences in EPD variability did not appear to be due to heritability. Crews et al. (2003) showed that carcass traits tended to have higher estimates of phenotypic variance than their corresponding live animal indicators. Wilson (1992) suggested that the expression of genetic differences for yearling RTU body composition traits may be limited by nutritional management.
|
Table 2 also contains a summary of EPD accuracy from the different models. Within each trait, accuracy increased from approximately 0.30 with L, through about 0.40 (0.35 to 0.42) with C, to a high of approximately 0.45 with F. Differences in mean accuracy among these models was partially due to differences in relative amounts of data. As indicated in Table 1, there were relatively more carcass data than live animal data. Considering that yearling live animal data are generally less costly to collect on a wider sample of animals, it can be predicted that only breeds that strongly encourage carcass data collection through organized progeny tests will have carcass databases that will be comparable in size to live animal databases. In populations for which live animal data are greatly more available than are carcass data, it would be expected that EPD based on L would have higher mean accuracy, especially among sires with no progeny with carcass data. However, because of the preference for expressing EPD on a carcass trait basis, this would not be equivalent to higher accuracy of carcass merit evaluation. The increased mean accuracy obtained with F compared to the more reduced models can be attributed to differences in amount of data as well. Because the reference sires were chosen on the basis of their links to either carcass or live data, minimum accuracy was near zero and maximum accuracy was similar for all traits. Even though bulls were required to have at least 15 progeny with either live or carcass data, few bulls had more than 50 progeny with either type of data in this analysis. These results support the intuitive assumption that F would be optimum because it includes all available data and estimates EPD on the desired carcass trait basis. The amount by which F may be considered superior to more reduced models will be dependent on accurate estimation of variance components and computing costs associated with a higher order model.
Correlations of EPD.
Correlations among EPD from L, C, and F are reported in Table 3. The correlation of SWT-L, RFAT-L, RREA-L, and RIMF-L with HCW-C, FAT-C, REA-C, and MAR-C were 0.31, 0.34, 0.26, 0.35, respectively. Although the correlations of EPD from L with those from F were higher (0.53 to 0.65) than those between L and C, these results suggest that choice of either reduced model will have a significant impact on sire rankings. The EPD from L and C were based on mutually exclusive sources of data; therefore, the expected value of their correlation is a function of EPD accuracy and genetic correlation. For example, with perfect accuracy, EPD from L and C would have simple correlations equal to the genetic correlation between the live and carcass traits (Henderson, 1984). Crews and Kemp (2002) showed the same general results and noted that EPD from F tended to be most highly correlated with EPD from C, except in the case of weight, for which they noted that yearling weight data was much more plentiful than carcass weight. Correlations of EPD from C with those from F were highest, ranging from 0.87 to 0.95. The similarities between EPD from C and F would be at least partially due to the common base on which these EPD are computed. Conversely, EPD from L are not expressed on a carcass basis, reducing the correlation of EPD from L with those from either C or F.
|
contains summary statistics for carcass traits of the 824 steers and heifers used to examine the association of progeny phenotype with sire EPD. Generally, mean and variability measures among the carcass records used to compute EPD and those reported in Table 4 were similar. Comparison indicates that among the data used to compute EPD, carcasses tended to have lower marbling score (5.04 vs. 5.24) compared to the data added after the study of Crews et al. (2003) although these differences may be considered minor.
|
. Mean, minimum, maximum, and SD of sire EPD follow similar trends as previously presented for the 202 reference sires. As discussed previously, regressions of progeny phenotype on EPD from model L would not generally have an expected value of 1.0 because of differences due to age at measurement, gender, and differences in genetic parameters between the live and carcass traits. Therefore, genetic regression was used to scale EPD based only on live data to a carcass basis. Genetic regression coefficients were 0.66, 1.72, 0.97, and 1.31 to predict HCW-R, FAT-R, REA-R, and MAR-R EPD using SWT-L, RFAT-L, RREA-L, and RIMF-L EPD.
|
contains model information and estimated partial regression coefficients for sire EPD. Model R2 values generally ranged from 0.46 to 0.59, with the exception of EPD related to fat thickness, for which model R2 values were lower, ranging from 0.32 to 0.34. Crews (2002) reported model R2 values that were either lower or comparable to those of the present study. In all cases, the regression of progeny phenotype on sire EPD was greater (P < 0.005) than 0, indicating a positive association between sire EPD and progeny phenotypes.
|
The prediction of carcass basis EPD using L and genetic regression (i.e., HCW-R, FAT-R, REA-R, and MAR-R) tended to improve the association with progeny phenotypes for carcass traits except carcass weight. The regression of progeny carcass fat thickness on sire FAT-R EPD was 1.00 ± 0.21, compared to the regression of 1.72 ± 0.35 obtained for sire RFAT-L EPD. There was also a numerical tendency for the regression coefficient of progeny marbling score on sire MAR-R EPD to be closer to 1.0 (0.97 ± 0.21) than that on sire RIMF-L EPD (1.27 ± 0.27). For carcass weight, however, the regression coefficient for sire HCW-R EPD was larger (P < 0.05) than 1.0 (1.63 ± 0.29). This indicates that HCW-R EPD was poorer at describing carcass weight differences in progeny than other EPD in this study. It is possible that this was due to poor estimation of the genetic covariance between SWT and HCW, which is the numerator of the genetic regression coefficient used to predict HCW-R from SWT-L EPD. Based on the report of Crews et al. (2003), the genetic covariance of SWT with HCW was 431.0 kg and the genetic variance of SWT was 655.4 kg2, resulting in a genetic regression of 0.66. Although not reported in tabular form, further investigation indicated that multiplication of SWT-L EPD by a genetic regression of 1.07 would have resulted in HCW-R EPD with a nearly perfect linear association with progeny carcass weight. Underestimation of the genetic covariance between SWT and HCW or overestimation of the genetic variance for SWT would have resulted in an underestimated genetic regression. These results generally imply that close linear associations exist between carcass phenotypes of progeny of Simmental bulls and sire carcass EPD based on combinations of live and carcass data.
Implications
Beef carcass expected progeny differences based on the combination of live animal and carcass data had a larger range and were more accurate for a larger sample of animals than carcass expected progeny differences based on live or carcass data alone. Sire expected progeny differences based on only yearling live animal data differed from those based on only carcass data and from those based on both types of data due at least partially to differences in age at measurement, gender, and variance components. Carcass phenotypes of progeny exhibited a close linear relationship with sire expected progeny difference based on both carcass data alone and a combination of live animal and carcass data. The linear association between carcass phenotype of progeny and sire expected progeny difference based on live animal data tended to be improved by scaling the live data expected progeny difference to a carcass trait basis. However, because of increased mean accuracy, expected progeny differences based on both live animal and carcass data would be considered optimal for genetic evaluation of beef carcass merit.
Footnotes
1 AAFC-LRC contribution number 38703049. Funding support provided by the American Simmental Association, and the AAFC Matching Investment Initiative is gratefully acknowledged. 
2 Correspondence: 5403 1st Avenue South (phone: 403-317-2288; fax: 403-382-3156; e-mail: dcrews{at}agr.gc.ca).
Received for publication July 21, 2003. Accepted for publication October 31, 2003.
Literature Cited
Basarab, J. A., D. Milligan, and J. Stitt. 1994. Relationship between expected progeny differences of Canadian Hereford sires and performance of their progeny in commercial herds. Can. J. Anim. Sci. 74:555–558.
Cameron, N. D. 1997. Selection Indices and Prediction of Genetic Merit in Animal Breeding. CABI, New York.
Crews, D. H., Jr. 2002. The relationship between beef sire carcass EPD and progeny phenotype. Can. J. Anim. Sci. 82:503–506.
Crews, D. H., Jr., and R. A. Kemp. 2001. Genetic parameters for ultrasound and carcass measures of yield and quality among replacement and slaughter beef cattle. J. Anim. Sci. 3008–3020.
Crews, D. H., Jr., and R. A. Kemp. 2002. Genetic evaluation of carcass yield using ultrasound measures on young replacement beef cattle. J. Anim. Sci. 80:1809–1818.
Crews, D. H., Jr., E. J. Pollak, R. L. Weaber, R. L. Quaas, and R. J. Lipsey. 2003. Genetic parameters for carcass traits and their live animal indicators in Simmental cattle. J. Anim. Sci. 81:1427–1433.
Devitt, C. J. B, and J. W. Wilton. 2001. Genetic correlation estimates between ultrasound measurements on yearling bulls and carcass measurements on finished steers. J. Anim. Sci. 79:2790–2797.
Henderson, C. R. 1984. Applications of Linear Models in Animal Breeding. University of Guelph, Guelph, Ontario.
Reverter, A., D. J. Johnston, H.-U. Graser, M. L. Wolcott, and W. H. Upton. 2000. Genetic analyses of live-animal ultrasound and abattoir carcass traits in Australian Angus and Hereford cattle. J. Anim. Sci. 78:1786–1795.
Van Vleck, L. D. 1992. Selection Index and Introduction to Mixed Model Methods. CRC Press, Boca Raton, FL.
Wilson, D. E. 1992. Application of ultrasound for genetic improvement. J. Anim. Sci. 70:973–983.[Abstract]
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
