The achievement of a given carcass specification is under moderate genetic control in cattle

Abstract

The objective of the present study was to estimate the genetic parameters associated with the achievement of desirable weight, conformation, and fat specifications, represented by a series of binary traits. The desired specifications were those stipulated by Irish beef processors, in accordance with the EUROP carcass grading system, and were represented by a carcass weight between 270 and 380 kg, a fat score between 2+ and 4= (between 6 and 11 on a 15-point scale), and a conformation score of O= or better (≥5 on a 15-point scale). Using data from 58,868 beef carcasses, variance components were estimated using linear mixed models for these binary traits, as well as their underlying continuous measures. Heritability estimates for the continuous traits ranged from 0.63 to 0.73; heritability estimates for the binary traits ranged from 0.05 to 0.19. An additional trait was defined to reflect if all desired carcass specifications were met. All genetic correlations between this trait and the individual contributing binary traits were positive (0.38 to 0.87), while all genetic correlations between this trait and the continuous carcass measures were negative (−0.87 to −0.07). The genetic parameters estimated in the present study signify that potential exists to breed cattle that more consistently achieve desirable carcass metrics at harvest.

Introduction

Beef carcass appraisal in Europe comprises measurements of carcass weight (CW), carcass conformation (CC), and carcass fat (CF) cover (DAFM, 2004), with the latter two being assessed on a linear scale (Pabiou et al., 2011). The classification of carcasses on such a scale provides the basis for payment to producers, with greater monetary incentives provided for the supply of carcasses with desirable yield and quality metrics. Desirable metrics are stipulated by beef processors in accordance with market signals from consumers, via the wholesalers, as well as the direct processing cost incurred per individual carcass (Kristensen et al., 2014). Therefore, the ability to consistently produce beef carcasses with optimum yield and quality metrics is highly desirable. From a cross-sectional analysis of a large national dataset, Kenny et al. (2020) documented that the majority of Irish cattle fail to achieve the desired specifications at harvest, with only 29% of bulls, 50% of heifers, and 39% of steers, harvested under 30 mo of age, achieving all of the desired CW, CC, and CF specifications.

As carcass output is the main revenue source in cattle production systems, the inclusion of carcass traits is commonplace in beef breeding indices (Amer et al., 1997; Barwick and Henzell, 2005; Berry et al., 2019a). Evaluations are generally based on quantitative measures of CW, CC, and CF, with the economic weights within the breeding objective also on a linear scale (Amer et al., 1998, 2001). This approach is adopted despite both CW and CF score having intermediate optimum values. Nonetheless, previous research in cattle has documented heritability estimates of between 0.08 and 0.58, between 0.24 and 0.46, and between 0.16 and 0.44 for the continuous traits of CW, CC, and CF, respectively (Pabiou et al., 2011; Kause et al., 2015; Englishby et al., 2016).

While the nongenetic factors associated with whether or not a beef carcass achieves a predefined, desired carcass specification are known (Kenny et al., 2020), no genetic parameter estimates have, to the best of our knowledge, been documented for such traits. A key objective of the present study, therefore, was to quantify the genetic contribution associated with the achievement of the desired carcass specifications and to determine the potential to breed cattle that more consistently achieve desirable carcass metrics at harvest. Should genetic variability be detected, breeding programs could form a useful component of a strategy to increase the proportion of carcasses that achieve the desired characteristics.

Materials and Methods

The data used in the present study were obtained from a preexisting database managed by the Irish Cattle Breeding Federation (ICBF; Bandon, Co. Cork, Ireland). Therefore, it was not necessary to obtain animal care and use committee approval in advance of conducting the study.

Data

Carcass information was available for 14,075,823 cattle harvested between the years 2003 and 2017, inclusive. The majority of animals included in the dataset were some crossbred combinations of the following breeds: Angus, Aubrac, Belgium Blue, Blonde d’Aquitaine, Charolais, Hereford, Holstein-Friesian, Jersey, Limousin, Saler, Shorthorn, and Simmental. All cattle births and inter-location movements are recorded by law; thus, the herd of birth, any subsequent herd the animal resided in, and the abattoir of harvest were known for each animal.

Using a similar method to that outlined by Ring et al. (2018a), the birth herd of each animal in the dataset was classified as either beef or dairy based on the average dam breed composition of the herd and the national frequency distribution of mean herd breed composition. If a herd had an average dam breed composition of ≤0.65 dairy breeds (i.e., Holstein-Friesian or Jersey), then the herd was classified as a beef herd; whereas if the average dam breed composition was >0.75 dairy breeds, the herd was considered to be a dairy herd. Any animals in the dataset born in herds that remained unclassified as beef or dairy were removed from the study. Only carcass records from heifers, steers, and young bulls harvested between 13 and 36 mo of age, with CWs between 100 and 800 kg, were retained for the current study. Any cattle with more than two inter-location movements during their lifetime, or a recorded inter-location movement within 100 d of harvest, were discarded. Additionally, any animal born from a dam with a parity number > 10, or from embryo transfer, was removed from the study. Furthermore, the sire and dam of all animals had to be recorded. Following these edits, 3,403,619 records remained.

General heterosis coefficients and recombination loss coefficients were calculated and classified for all animals using the methods previously outlined by Kenny et al. (2020). Furthermore, contemporary groups of finishing herd, year, season, and sex were formed using parameters and an algorithm described previously by Kenny et al. (2020). Only contemporary groups with at least 10 records were retained. Following all edits, carcass data for 1,493,431 cattle remained, of which 639,704 were steers, 427,642 were heifers, and 426,085 were bulls. The sample dataset contained 88,904 contemporary groups, representing 16,801 different herds.

Trait definition

Three carcass metrics, reflecting the yield and quality of the carcasses, were available for all animals in the study, namely CW, CC, and CF. The measurement of (cold) CW reflects the kilograms of saleable beef on which the supplier is eligible to receive payment. CC depicts the development of the carcass with a particular emphasis on the round, back, and shoulders, while CF is a measurement of the level of external fat on the carcass, as well as the level of fat within the thoracic cavity (Conroy et al., 2010). The latter two carcass metrics, which are assessed by video-image analysis (Pabiou et al., 2010), are used to calculate the premiums or deductions applicable per kilogram of CW. Both CC and CF are scored on 15-point linear scales, in accordance with the EUROP grading system. The CC score is represented by the letters E (best), U, R, O, and P (worst), whereas CF is scored from 1 (thin) to 5 (fat), with each CC and CF score divided into three subdivisions (–, =, and +).

The desired specifications for CW, CC, and CF used in the current study were derived from those stipulated by wholesalers and Irish beef processors at the time the study was carried out. The desired carcass specifications were represented by a CW between 270 and 380 kg, a CC score of O= or better (≥5 on a 15-point numeric scale), and a CF score between 2+ and 4= (6 to 11 inclusive on a 15-point numeric scale). Whether or not an individual animal fulfilled the desired specification(s) was defined as a series of binary traits. Binary traits were firstly created to represent the success or failure of an animal in achieving each individual desired specification separately (i.e., the CW, CC, or CF specification). Binary traits were also created to represent the success or failure of animals achieving different combinations of the individual desired specifications, including the achievement of all the desired specifications (i.e., simultaneous achievement of the desired CW, CC, and CF specifications). Additionally, binary traits were generated to represent whether an individual animal achieved the minimum desired thresholds for CW (i.e., ≥270 kg) or CF (i.e., a EUROP fat score ≥2+). The continuous carcass traits for CW, CC, and CF were also available for analysis. The distribution of CWs achieved by cattle in the present study is in Supplementary Figure S1, with the distribution of CC scores and CF scores in Supplementary Figure S2.

For an additional series of analyses, a new series of continuous traits were created to reflect the deviation in the actual CW, CC, and CF scores of an animal from the respective desired threshold values. Animals that achieved the desired specification for the carcass trait under consideration were assigned a value of zero. Where the value of the carcass trait was lower than the minimum threshold, the deviation trait was calculated as the carcass metric achieved minus the minimum desired threshold; where the value of the carcass trait was higher than the maximum threshold, the deviation trait was calculated as the maximum desired threshold minus the carcass metric achieved. Therefore, all deviation traits were negative.

(Co)variance component estimation

Prior to the estimation of variance components, a random subset of the contemporary groups of cattle harvested between the years 2012 and 2015 was created; the random subset comprised of 3,637 contemporary groups, which contained a total of 58,868 cattle (17,890 bulls, 17,275 heifers, and 23,703 steers) harvested from 2,923 herds.

Genetic and residual variance components were estimated for all traits using linear mixed models in ASREML 3.0 (Gilmour et al., 2009); Pabiou et al. (2011) failed to detect any great difference in variance components for carcass traits by gender using Irish data and thus all genders were considered in the analysis, albeit with an adjustment for mean gender differences via their inclusion in the contemporary group. Fixed effects considered in all models included contemporary group, dam parity, birth herd type (i.e., beef or dairy), whether the animal was born a singleton or a twin, and heterosis and recombination loss coefficient classes. A direct additive genetic effect (N(0,Aσa²)) was included as a random term, alongside a random residual term N(0,Iσe²), where σa² is the additive genetic variance, σe² is the residual variance, A is the numerator relationship matrix, and I is an identity matrix. Phenotypic and genetic correlations between the traits of interest were calculated using a series of bivariate sire linear mixed models. The pedigree of each animal/sire was traced back to a founder population, which were allocated to genetic groups by breed; the pedigree file contained a total of 4,098,259 animals.

Genetic evaluation

Estimated breeding values (EBVs) and the respective reliabilities were generated for all binary specification traits, as well as for the traditional continuous traits of CW, CC, and CF using the Mix99 software suite (Lidauer et al., 2017). All genetic evaluations were univariate and utilized the variance components estimated in the present study. Prior to conducting the genetic evaluations, the phenotypic data for all 186,238 cattle (53,824 bulls, 56,424 heifer, and 75,990 steers) harvested in 2017 were masked. Therefore, EBVs were calculated for the 1,307,193 cattle harvested between the years 2003 and 2016 based on their phenotype and pedigree, with the breeding values for cattle harvested in 2017 estimated through their relationships, via the numerator relationship matrix, with the phenotyped animals. All breeding values were expressed relative to those of a base population set to be 160 Holstein-Friesian sires born between the years 2005 and 2010, with more than 100 progeny harvested before the 2017 calendar year.

Parental average EBVs for the continuous carcass traits and binary specification traits of all validation animals harvested in 2017 were stratified into 10 groups, each based on the respective EBVs of interest. Only cattle with a reliability ≥ 0.10 for the trait of interest were retained. CW, CC, and CF least squares means were calculated using linear mixed models in PROC MIXED in SAS 9.4 for each EBV stratum derived from the genetic evaluation using either the continuous or binary traits. Furthermore, using multivariable logistic regression models and a logit link function in SAS 9.4, the predicted probability of achieving each carcass specification was calculated for cattle within each of the different EBV strata. The probability of achieving the specification was calculated for a reference animal that was set to be a singleton beef-origin steer, born from a third parity dam, and harvested at 20 mo of age, with 0% heterosis and 0% recombination loss. Fixed effects considered in both the linear and logistic models were dam parity, whether the animal was born as a singleton or twin, heterosis classes, recombination loss classes, and the EBV stratum for the trait of interest, as well as a three-way interaction between birth herd type (i.e., beef or dairy), sex (i.e., bull, heifer or steer), and age at harvest (in months). Contemporary group (i.e., finishing-herd-year-season-sex) was included as a random term in all models.

To evaluate the accuracy of the different genetic evaluations in predicting whether a carcass achieved the intermediate optimum CW or CF specifications, relative operating characteristic (ROC) curves were generated, with the accuracy of the predictions determined from the area under the curves (Hajian-Tilaki, 2013). All ROC curves were generated from multivariable regression models using a logit link function in SAS 9.4. The models included a variable representing the EBV stratum from the binary specification trait or the continuous carcass trait, as well as sex (i.e., bull, heifer, or steers) and birth herd type (i.e., beef or dairy) as independent variables. The binary dependent variable in all models was a variable representing whether or not an animal achieved the desired CW specification of 270 to 380 kg or the desired CF specification of 2+ to 4=. The area under the ROC curves were calculated and compared using the methods outlined by DeLong et al. (1988).

Results

Variance components

The phenotypic mean, genetic standard deviation, and heritability estimates for all carcass traits are given in Table 1. The genetic standard deviation for the continuous measures of CW, CC, and CF was 26.16 kg, 1.01 units, and 0.94 units respectively, while the heritability estimates for the continuous carcass measures were between 0.63 and 0.73. In comparison, the heritability of the binary traits ranged from 0.05 (minimum weight specification) to 0.19 (trait representing the simultaneous achievement of the CC and CW specifications), with the genetic standard deviation of the traits ranging from 0.05 to 0.17 units. The heritability estimates for the CW, CC, and CF deviation traits were 0.95, 0.16, and 0.27, respectively. Additionally, heritability estimates of 0.74 and 0.66 were calculated for continuous weight deviation traits, in which the CW thresholds were set to be 300 to 350 kg and 315 to 345 kg, respectively.

Table 1.

Means, genetic standard deviations, and heritability estimates for the quantitative measures of the carcass metrics, the desired specification traits for various combinations of the CW (270 to 380 kg), CC (≥ O= conformation score on the EUROP carcass classification scale), and CF (2+ to 4= fat score on the EUROP carcass classification scale) specifications, and the continuous deviations from the desired specification thresholds

Trait, unit	Mean	Genetic standard deviation	Heritability (standard error)
Quantitative measure:
CW, kg	350.22	26.16	0.68 (0.02)
CC1	7.93	1.01	0.73 (0.02)
CF1	7.63	0.94	0.63 (0.02)
Binary traits:
CW	0.62	0.16	0.18 (0.02)
CC	0.84	0.08	0.12 (0.01)
CF	0.84	0.11	0.11 (0.01)
Minimum CW	0.92	0.05	0.05 (0.01)
Minimum CF	0.88	0.09	0.10 (0.01)
CW and CC	0.50	0.17	0.19 (0.02)
CW and CF	0.52	0.14	0.12 (0.01)
CC and CF	0.71	0.14	0.14 (0.01)
CW, CC, and CF	0.42	0.17	0.17 (0.02)
Continuous deviation traits:
CW, kg	−14.69	18.80	0.95 (0.02)
CC1	−0.24	0.16	0.16 (0.01)
CF1	−0.15	0.22	0.27 (0.02)

¹Scored on a scale from 1 (lowest) to 15 (highest) in accordance with the EUROP grading system.

Correlations

The phenotypic and genetic correlations among the various specification traits are given in Table 2. All phenotypic and genetic correlations among the specification traits were positive, with the exception of the genetic correlation between the minimum CW specification trait and the intermediate optimum CF specification trait. Both the phenotypic and genetic correlations among the desired CW (270 to 380 kg), CC (≥ O=), and CF (2+ to 4=) specifications were weak (i.e., ≤0.12). Furthermore, the minimum desired specification for CW (≥ 270 kg) and CF (≥ 2+) had a phenotypic correlation with their intermediate optimum equivalents of 0.54 and 0.84, respectively, as well as respective genetic correlations of 0.06 and 0.60.

Table 2.

Phenotypic correlations¹ (above diagonal) and genetic correlations (below diagonal, standard errors in parenthesis) among the various combinations of desired specifications for CW (270 to 380 kg), CC (≥ O= conformation score), and CF ( 2+ to 4= fat score)

	CW	CC	CF	Min CW	Min CF	CW & CC	CW & CF	CC & CF	CW, CC & CF
CW		0.03	0.01	0.54	0.02	0.82	0.76	0.01	0.65
CC	0.12 (0.06)		0.04	0.11	0.06	0.48	0.03	0.55	0.40
CF	0.05 (0.08)	0.06 (0.07)		0.06	0.84	0.01	0.50	0.78	0.42
Minimum CW	0.06 (0.08)	0.29 (0.07)	−0.14 (0.10)		0.10	0.38	0.39	0.07	0.29
Minimum CF	0.16 (0.07)	0.04 (0.07)	0.60 (0.05)	0.03 (0.10)		0.01	0.40	0.63	0.31
CW and CC	0.93 (0.01)	0.42 (0.04)	0.04 (0.07)	0.37 (0.07)	0.07 (0.07)		0.63	0.26	0.81
CW and CF	0.94 (0.01)	0.11 (0.06)	0.29 (0.07)	0.17 (0.09)	0.28 (0.07)	0.85 (0.02)		0.38	0.85
CC and CF	0.14 (0.06)	0.61 (0.04)	0.81 (0.02)	0.17 (0.08)	0.60 (0.05)	0.22 (0.06)	0.43 (0.06)		0.59
CW, CC, and CF	0.87 (0.02)	0.43 (0.05)	0.38 (0.07)	0.35 (0.07)	0.29 (0.07)	0.95 (0.01)	0.93 (0.01)	0.45 (0.05)

¹Standard errors were ≤0.005.

The phenotypic and genetic correlations between the continuous measures of CW, CC, and CF and the various binary specification traits are given in Table 3. The phenotypic correlations between CW, CC, and CF and their respective binary carcass specification traits were −0.32, 0.36, and 0.06, while the genetic correlations between the continuous measures and their corresponding binary specification traits were −0.71, 0.44, and −0.06, respectively. Additionally, the genetic correlation between the continuous CW trait and the minimum CW specification was 0.75, while the genetic correlation between the continuous CF trait and the minimum CF specification was 0.76. All phenotypic correlations between the deviation traits and the binary specification traits were positive (0.003 to 0.78), while the genetic correlations between the deviation traits and the binary specification traits ranged from −0.13 to 0.94 (Table 4).

Table 3.

Phenotypic correlations¹ and genetic correlations (standard errors in parenthesis) between the quantitative measurements for CW, CC, and CF and the various combinations of the CW (270 to 380 kg), CC (≥ O= conformation score), and CF (2+ to 4= fat score) specifications

	Phenotypic			Genetic
Trait	CW	CC	CF	CW	CC	CF
Quantitative measure:
CW	—	0.46	0.14	—	0.56 (0.02)	−0.49 (0.03)
CC	—	—	−0.04	—	—	−0.25 (0.03)
Desired specification:
CW	−0.32	−0.17	0.01	−0.71 (0.03)	−0.43 (0.04)	0.10 (0.05)
CC	0.13	0.36	0.06	0.11 (0.04)	0.44 (0.04)	−0.01 (0.05)
CF	0.07	0.02	0.06	−0.14 (0.06)	−0.01 (0.05)	−0.06 (0.06)
Minimum CW	0.34	0.15	0.11	0.75 (0.04)	0.43 (0.06)	0.17 (0.06)
Minimum CF	0.17	0.01	0.36	−0.08 (0.06)	−0.16 (0.05)	0.76 (0.04)
CW and CC	−0.22	−0.02	−0.01	−0.21 (0.03)	−0.36 (0.03)	−0.01 (0.04)
CW and CF	−0.24	−0.13	0.01	−0.70 (0.03)	−0.39 (0.03)	0.002 (0.05)
CC and CF	0.10	0.18	0.04	−0.08 (0.05)	0.12 (0.05)	−0.09 (0.05)
CW, CC, and CF	−0.22	−0.02	0.00	−0.64 (0.03)	−0.32 (0.04)	−0.07 (0.05)

¹All standards errors were ≤0.006.

Table 4.

Phenotypic correlations¹ and genetic correlations (standard errors in parenthesis) between the continuous deviation from the desired specification thresholds, the continuous measurements for CW, CC, and CF, and the various combinations of the CW (270 to 380 kg), CC (≥ O= CC score), and CF (2+ to 4= fat score) specifications

	Continuous deviation traits
	Phenotypic			Genetic
Trait	CW	CC	CF	CW	CC	CF
Continuous deviation traits:
CW	—	0.06	0.04	—	0.06 (0.04)	−0.12 (0.04)
CC	—	—	0.06	—	—	0.10 (0.05)
Quantitative measures:
CW	−0.57	0.14	0.04	−0.77 (0.01)	0.07 (0.04)	−0.14 (0.04)
CC	−0.26	0.40	0.02	−0.39 (0.02)	0.29 (0.03)	0.06 (0.04)
CF	0.01	0.07	−0.15	−0.01 (0.03)	0.03 (0.04)	−0.40 (0.04)
Desired specification:
CW	0.53	0.05	0.03	0.84 (0.02)	0.10 (0.05)	0.30 (0.06)
CC	0.04	0.78	0.04	0.11 (0.04)	0.31 (0.08)	0.07 (0.06)
CF	0.01	0.05	0.74	−0.11 (0.05)	0.09 (0.06)	0.94 (0.02)
Minimum CW	0.20	0.14	0.05	0.20 (0.06)	0.25 (0.06)	−0.13 (0.09)
Minimum CF	0.003	0.07	0.49	0.10 (0.06)	0.08 (0.06)	0.38 (0.06)
CW and CC	0.46	0.38	0.02	0.77 (0.02)	0.45 (0.04)	0.03 (0.06)
CW and CF	0.40	0.04	0.38	0.81 (0.03)	0.11 (0.06)	0.35 (0.06)
CC and CF	0.01	0.43	0.59	0.14 (0.05)	0.64 (0.04)	0.75 (0.03)
CW, CC, and CF	0.37	0.28	0.31	0.73 (0.03)	0.36 (0.05)	0.27 (0.06)

¹All standard errors were ≤0.006.

Genetic selection for CW and the desired weight specification

When the genetic evaluation was based on the continuous CW trait, the least squares mean for CW increased consistently from 304.7 kg for cattle in the poorest EBV stratum to 355.0 kg for cattle in the highest EBV stratum (Figure 1). In addition to carcasses getting heavier with increasing EBV stratum for the continuous CW trait, the predicted probability of a 20-mo-old beef-origin steer achieving a CW > 380 kg increased consistently from 0.06 in the poorest EBV stratum to 0.62 in the highest EBV stratum. Furthermore, for the continuous CW trait, the predicted probability of a beef-origin steer harvested at 20 mo of age having a CW between 270 and 380 kg increased from 0.73 in the poorest EBV stratum to 0.75 in the third poorest EBV stratum, before declining to 0.41 in the highest EBV stratum.

Figure 1.

Least squares means (95% confidence intervals as error bars) for the CWs (solid line, primary axis) achieved by groups of cattle with similar estimated breeding values for the binary weight specification trait (top graph) and for the continuous CW trait (bottom graph) as well as the predicted probability of 20-mo-old beef-origin steers achieving the desired weight specification (between 270 and 380 kg; dashed line, secondary axis).

When the validation cattle were ranked on their EBV for the binary CW specification trait, the CW least squares mean declined consistently from 351.1 kg in the poorest EBV stratum to 315.5 kg in the highest EBV stratum (Figure 1). Furthermore, the predicted probability of a 20-mo-old beef-origin steer exceeding a CW of 380 kg reduced consistently from 0.64 to 0.16 as the EBV for the binary weight specification trait increased, while concurrently the predicted probability of the CW being between 270 and 380 kg increased consistently from 0.38 to 0.78.

Genetic selection for CC score and the desired CC specification

In general, regardless of whether the genetic evaluation was based on the continuous CC trait or the binary CC specification trait, the least squares mean for CC score as well as the predicted probability of having a CC score ≥ O= increased with each EBV stratum (Figure 2). When the genetic evaluation was based on the continuous CC trait, the least squares mean for CC score (measured on a 15-point linear scale) increased consistently from 4.50 for cattle in the poorest EBV stratum to 9.26 for cattle in the highest EBV stratum. Furthermore, for the continuous CC trait, the predicted probability of a 20-mo-old beef-origin steer achieving the desired CC specification increased from 0.22 in the poorest EBV stratum to 0.99 in the highest EBV stratum.

Figure 2.

Least squares means (95% confidence intervals as error bars) for the CC scores (solid line, primary axis) achieved by groups of cattle with similar estimated breeding values for the binary CC specification trait (top graph) and for the continuous CC trait (bottom graph) as well as the predicted probability of 20-mo-old beef-origin steers achieving the desired CC specification (≥5 on a 15-point numeric scale; dashed line; secondary axis).

When the validation cattle were ranked on their genetic merit for the binary CC specification trait, the least squares mean for CC score increased from 4.46 for cattle in the poorest EBV stratum to 7.86 for cattle in the second highest EBV stratum but decreased slightly to 7.61 for cattle in the highest EBV stratum. Additionally, for the binary CC specification trait, the predicted probability of a 20-mon-old beef-origin steer having a CC ≥ O= (≥5 on a 15-point scale) increased from 0.29 in the poorest EBV stratum to 0.99 in the second highest EBV stratum, before declining to 0.97 in the highest EBV stratum (Figure 2).

Genetic selection for CF score and the desired CF specification

From the poorest EBV stratum to the highest EBV stratum of the continuous CF trait, the least squares mean for CF score (measured on a 15-point linear scale) increased consistently from 6.03 to 9.75 (Figure 3). Further to the carcasses getting increasingly fatter as the EBV for the continuous CF trait increased, the predicted probability of a 20-mo-old beef-origin steer having a CF score between 2+ and 4= increased from 0.69 in the poorest EBV stratum to 0.90 in the sixth poorest EBV stratum, before declining thereafter to 0.67 in the highest EBV stratum.

Figure 3.

Least squares means (95% confidence intervals as error bars) for the CF scores (solid line, primary axis) achieved by groups of cattle with similar estimated breeding values for the binary CF specification trait (top graph) and for the continuous CF trait (bottom graph), as well as the predicted probability of 20-mo-old beef-origin steers achieving the desired CF specification (between 6 and 11 on a 15-point numeric scale; dashed line, secondary axis).

On the other hand, the least squares mean for CF score increased consistently from 6.37 in the poorest EBV stratum of the binary CF specification trait to 8.16 in the highest EBV stratum of the binary CF specification trait (Figure 3). Furthermore, for the binary CF specification trait, the predicted probability of a beef-origin steer harvested at 20 mo of age achieving the desired CF specification of 2+ to 4= (between 6 and 11 on a 15-point scale) increased consistently from 0.69 in the poorest EBV stratum to 0.91 in the highest EBV stratum.

Comparison of predictions from the continuous and binary breeding values

The area under the ROC curves generated from the EBV strata of the continuous CW trait and binary CW specification trait are given in Table 5. Based on the area under the curves being greater than 0.5 (P < 0.001), the parental average breeding values of either the continuous CW trait or the binary CW specification trait can distinguish, with some level of accuracy, between cattle that are expected to achieve the desired CW specification and those that are not. Moreover, the inclusion of additional predictors in the multiple regression models, such as birth herd type (i.e., beef or dairy) or sex (i.e., bull, heifer or steer), further improved the accuracy of the predictions (i.e., increased the area under the curve; P < 0.001) over and above just the inclusion of the variable representing the EBV strata. No difference (P = 0.065) existed between the area under the ROC curve generated from just the EBV stratum of the binary CW specification trait and the area under the ROC curve generated from just the EBV stratum of the continuous CW trait. In the multiple regression models using birth herd type or sex as additional predictors, the models exploiting the EBV stratum from the binary CW specification trait outperformed (P < 0.01; Table 5) the corresponding models that used the EBV stratum from the continuous CW trait. On the other hand, when both birth-herd type and sex were included as fixed effects in the multiple regression models, the area under the ROC curve generated from the model exploiting the EBVs of the continuous CW trait was greater than that from the model exploiting the EBVs of the binary CW specification trait (P = 0.003; Table 5).

Table 5.

Areas under the ROC curves¹ generated using the parental average EBV strata from either the continuous carcass trait of interest or the binary specification trait of interest, as well as birth herd type (i.e., beef or dairy) and sex (i.e., bull, heifer, and steers), to predict whether an animal achieved the desired CW specification (i.e., 270 to 380 kg) or the desired CF specification (i.e., 2+ to 4=)

	Parental average breeding value
Predictor(s)	Binary	Continuous	P-value2
CW specification
EBV stratum	0.685	0.683	0.065
EBV stratum and birth herd type	0.697	0.694	0.007
EBV stratum and sex	0.723	0.721	0.001
EBV stratum, birth herd type, and sex	0.731	0.733	0.003
CF specification
EBV stratum	0.577	0.563	<0.001
EBV stratum and birth herd type	0.693	0.687	<0.001
EBV stratum and sex	0.577	0.576	0.644
EBV stratum, birth herd type, and sex	0.695	0.691	<0.001

¹All areas under the curve were significantly different from 0.5 (P < 0.001).

²Denotes the significance of the difference between the area under the curve using the binary breeding values and those under the curve using the continuous breeding values.

Similar to CW, the area under the ROC curves generated from the EBV strata of the continuous CF trait and binary CW specification trait (Table 5) signify that the parental average EBVs can be used to predict, with some level of accuracy, whether cattle will achieve the desired CF specification. With the exception of the model using the continuous CF EBV stratum and birth herd type as predictors (P = 0.19), when compared with the models that included only the EBV stratum, the inclusion of sex and/or birth herd type as additional predictors in the multiple regression model increased (P < 0.001) the area under the curves. Apart from the multiple regression models including the EBV stratum and birth herd type as fixed effects (P = 0.64), the area under the curves generated from the models exploiting the EBVs of the binary CF specification trait were greater than those exploiting the EBVs of the continuous CF trait (P < 0.001 Table 5).

Discussion

The characteristics of beef carcasses in many countries are conditioned by the goldilocks principle of “just the right amount”, especially in relation to CW and CF cover. The underlying justification for such stipulations have been discussed elsewhere (Kenny et al., 2020), but mainly revolve around consumer preferences for a specific carcass cut size (Maples et al., 2018) and fat cover (Frank et al., 2017), as well as the processing costs per individual carcass (Kristensen et al., 2014). This has prompted the enforcement, through price penalties, of carcass specifications for CW, CF score, and a minimum CC score in Ireland and some of Europe. The fact that the majority (i.e., 57%) of carcasses in the present study failed to achieve at least one of these desired specifications is thus a concern for the future viability of the industry. While Kenny et al. (2020) documented some of the nongenetic animal-level risk factors associated with not achieving these specifications in Irish cattle, they also highlighted that some of these factors cannot be easily changed; for example, the generally lighter carcasses of progeny from dairy females cannot easily be negated by breeding on the dam side, without likely repercussions for genetic gain in dairy-specific attributes. Therefore, the objective of the current study was to quantify whether breeding programs could be used to increase the proportion of animals that achieve the enforced carcass specifications. In the dairy example, such a tool could be deployed via an index to rank beef bulls for use on dairy females (Berry et al. 2019a).

As CW and CF score have an intermediate optimum (i.e., the goldilocks principle), the current strategy of using EBVs based on a linear scale may not be sensible. This hypothesis was tested using a validation population comparing the status quo approach to a method where the linear scale was dichotomized based on whether or not the desired carcass credentials were achieved. The presence of genetic variation for these binary traits reflecting adherence to the specifications indicates the potential use of animal breeding as a strategy to improve the proportion of carcasses achieving the desired specifications.

Genetic parameters of the continuous carcass traits and the binary specification traits

The heritability estimates reported in the present study for CW, CC, and CF (i.e., 0.68, 0.73, and 0.63, respectively) were generally larger than those previously reported in cattle studies (i.e., 0.06 to 0.65, 0.04 to 0.46, and 0.00 to 0.40, respectively; Hickey et al., 2007; Pabiou et al., 2011). Moreover, the genetic standard deviations documented in the present study for CW (26.16 kg), CC (1.10 units), and CF (0.94 units) were also generally larger than those previously reported for cattle (6.63 to 25.16 kg, 0.20 to 0.73 units, and 0.00 to 0.95 units, respectively; Hickey et al., 2007; Pabiou et al., 2011). The differences in the genetic variation reported for the continuous carcass traits could be attributed to differences between the corresponding production systems of cattle investigated in the various studies, with the previous studies being confined to only steers and heifers (Pabiou et al., 2011), or only dairy-origin cattle (Hickey et al., 2007). Genotype by environment interactions have been well-documented in the past to contribute to differences in the estimates of genetic parameters (Berry et al., 2003a; Ibi et al., 2005; Ring et al., 2018b).

Although no variance components have been previously reported for the binary specification traits, the genetic standard deviation reported in the present study for these binary traits (0.05 to 0.17) was similar, albeit slightly higher, to that previously documented in dairy cattle for the binary fertility trait of pregnancy to first service (0.05; Berry et al., 2003b). Therefore, based on estimated genetic variation, the potential to improve the proportion of cattle that achieve desirable carcass specifications should be comparable to that of improving the reproductive performance of dairy cattle, which has been well-acknowledged to have been achieved (Berry et al., 2014).

Deployment

While the presence of genetic variability is important for genetic gain, the rate of genetic gain is also a function of selection intensity, accuracy of selection, and generation interval (Rendel and Robertson, 1950). Accuracy of selection is a function of the heritability of the trait and the quantity of phenotypic information available (Cassell, 2009; Berry et al., 2019b). The heritability in the present study of 0.05 to 0.19 for the binary traits implies that, to achieve an accuracy of selection of 0.70, in the absence of any genomic or parental information, phenotypic data on 20 to 76 progeny would be required. Therefore, the number of progeny records required to achieve a high accuracy is not large, particularly considering that carcass traits are not sex linked and, in Ireland at least, the data are routinely available on almost all animals harvested. Although not all animals will have a known sire or be from informative contemporary groups, achieving high accuracy of selection should not be a hindrance to genetic gain. Furthermore, the average age at harvest in the present study was 23.8 mo; hence, a relatively short generation interval should be achievable for a terminal index that includes such traits (Amer et al., 1997; Berry et al., 2019a). All in all, there is no reason why rapid genetic gain could not be achieved for the traits evaluated in the present study.

One of the main uses of the results from the present study will be in the back-end of sire advice systems. Carthy et al. (2019) used linear programming in their proposal for mating dairy bulls to dairy females in the pursuit of maximizing the level of genetic improvement while minimizing the level of inbreeding in potential progeny. The algorithm proposed by Carthy et al. (2019) could be easily expanded to consider the probability of the resulting progeny achieving each carcass specification given the genetic merit of both mates, as well as the cow-level factors associated with the achievement of the carcass specifications documented by Kenny et al. (2020). Consideration should also be given in the mating advice algorithm to the likelihood of a difficult calving, given that this is an important consideration when choosing bulls for use on dairy females (Berry et al., 2019c). The approach proposed could also be deployed for assortative mating of beef bulls to dairy females (Berry et al., 2019a) or for beef bulls to beef females. Alternatively, to increase the proportion of beef cattle with desirable carcass specifications, the binary specification traits could be considered as goal traits in beef breeding indexes, although the emphasis placed on the traits and their covariance with other traits in the breeding goal would first have to be considered. The inclusion of binary traits, such as non-return rate and survival, in cattle breeding indexes has been previously demonstrated in numerous studies (Boettcher and Van Doormaal, 1999; González-Recio and Alenda, 2005). Berry et al. (2019a) proposed an alternative approach where they incorporated the continuous carcass traits directly in a cattle breeding index but with a nonlinear economic value applied to the trait.

Furthermore, the predictive ability of achieving the desired specification based on genetic merit, coupled with other readily available information, such as birth herd type and sex (Kenny et al., 2020), could underpin a decision support tool for the purchase of cattle for finishing. While such a tool could provide information on the likelihood of an animal achieving the desirable carcass specifications, the ability to detect animals with a high probability of failing to achieve a specification could trigger some corrective measures within the feedlot to mitigate that probability. Such strategies could include castration of male calves (Nogalski et al., 2018), altering the level of concentrates fed during the finishing period, or altering the length of the finishing period (Sami et al., 2004).

Conclusions

The ability to improve the probability of progeny achieving a given carcass specification with an intermediate optimum and, by extension, the population as a whole, could be aided by breeding programs. Such breeding programs could be based on a breeding goal that includes a binary trait of whether or not a desired carcass specification is likely to be attained. The probability statistics could be generated for each male and female candidate parent for each trait separately and assortative mating, via sire advice, invoked to maximize the chances further. To be successful, market signals for the purchase of cattle should reflect the probability statistics, although abattoirs must be consistent in their application of price penalties, irrespective of the prevailing supply and demand conditions at the time of harvest.

Supplementary Data

Supplementary data are available at Journal of Animal Science online.

Supplementary Figure S1: Distribution of the CWs achieved by cattle in the present study, with dashed lines representing the desired thresholds of the weight specification.

Supplementary Figure S2: Distributions of the CC and CF scores achieved by cattle in the present study, with the dashed line representing the desired specification thresholds.

Glossary

Abbreviations

CC

carcass conformation

CF

carcass fat

CW

carcass weight

EBV

estimated breeding value

ROC

relative operating characteristic

Conflict of interest statement

The authors have no conflict of interest to declare.