Young to Old Ossification Beef Maturity

• Journal Listing
• J Anim Sci
• v.97(i); 2019 Jan
• PMC6313154

J Anim Sci. 2019 Jan; 97(one): 144–155.

Prediction of ossification from live and carcass traits in young beef cattle: model development and evaluation¹

Boyd W Gudex

^oneNSW Department of Primary Industries, Livestock Industries Heart, University of New England, Armidale, New South Wales, Australia

Malcolm J McPhee

¹NSW Section of Primary Industries, Livestock Industries Center, Academy of New England, Armidale, New S Wales, Australia

Victor H Oddy

ⁱNSW Department of Main Industries, Livestock Industries Heart, University of New England, Armidale, New South Wales, Commonwealth of australia

Brad J Walmsley

¹NSW Department of Principal Industries, Livestock Industries Centre, University of New England, Armidale, New Due south Wales, Australia

³Animate being Genetics and Breeding Unit, University of New England, Armidale, New S Wales, Australia

Received 2018 Jun fourteen; Accepted 2018 Oct 30.

Abstruse

Physiological maturity, measured as carcass ossification [10 unit increments (100, 110, 120, …)], is used by the United States Department of Agriculture and the Meat Standards Australia carcass grading systems to reflect historic period-associated differences in beef tenderness and decide producer payments. In most commercial cattle herds, the exact historic period of animals is unknown; thus, prediction of ossification in association with phenotypic prediction systems has the capacity to assist producer decision making to improve carcass and eating quality. This study developed and evaluated prediction equations that use either live animal or carcass traits to predict ossification for use in phenotypic prediction systems to predict meat quality. The average ossification in the model evolution dataset was 138 with a SD of 21 and a range between 100 and 200. Model development involved regressing diverse combinations of live beast traits: age at recording, sex, alive weight (BW), average daily gain, ultrasound scanned middle muscle surface area, 12/13th rib and subcutaneous P8 rump fat thickness; or carcass traits: age at slaughter, sex, hot standard carcass weight (HSCW), carcass center muscle surface area, marble score, rib, and P8 rump fatty (CP8) thickness, against ossification. The models were challenged with data from 3 independent datasets: one) Angus steers produced by divergent selection for visual musculus score; 2) temperate (Angus, Hereford, Shorthorn and Murray Grey) steers and heifers; and three) tropically adapted (Brahman and Santa Gertrudis) steers and heifers. V models with adjusted R ^two _adj higher up 0.55 were evaluated. When challenged with dataset 1, the absolute hateful bias (MB) and root mean square error of prediction (RMSEP) ranged from 0.i to 4.ii, and 9.8 to 10.7, which are within the bounds of the 10 signal increment on the ossification scale. When subsequently challenged with dataset 2, MB and RMSEP ranged from 2.eight to 13.4, and nineteen.6 to 23.7, respectively; and with dataset 3, MB and RMSEP ranged from 14.iv to 17.5, and 23.iii to 31.9, respectively. Mostly, when compared in relation to the ossification scale, all evaluated models had like accuracy. For predicting meat quality, the model containing live animal traits considered nigh useful was [85.35 + 0.16 × BW + x.94 × sex – 0.09 × sex × BW (adjusted R ^two = 0.59; SE = xiii.51)] and the about useful model containing carcass traits was [107.15 + 11.53 × sex + 1.10 × CP8 + 0.xvi × HSCW – 0.15 × sexual practice × HSCW (adjusted R ² = 0.60; SE = 13.39)].

Keywords: beef cattle, carcass, live animal, maturity, meat quality, ossification

INTRODUCTION

Market signals provided the accent during the 1990's and 2000's for the beef industry to increment the focus on improving carcass and meat quality (Johnston et al., 2003). This is embodied in carcass grading systems developed by the United States Section of Agriculture (USDA, 1997) and Meat Standards Australia (MSA, Watson et al., 2008). Carcass ossification is used by both the MSA and USDA carcass grading systems when predicting the eating quality of beef which impacts on the payments made to producers. Ossification is an assessment of the calcification of the spinous processes in the sacral, lumbar, and thoracic vertebrae (AUS-MEAT, 2005; Watson et al., 2008) and is used within grading systems as an assessment of physiological maturity. Physiological maturity is used to reflect age-associated differences in beef tenderness (Schonfeldt and Strydom, 2011). In live commercial cattle, exact age or other measures of physiological maturity are often unknown.

The predictions fabricated by carcass grading systems can be used in association with phenotypic prediction systems, such as the BeefSpecs calculator (Walmsley et al., 2014), to provide predictions of eating quality up to 200 d prior to slaughter. This has the capacity to assist producers make management decisions to improve compliance rates with carcass marketplace specifications and ameliorate eating quality. To achieve this, the ossification input required by the carcass grading systems could be predicted from the live animal [e.g., historic period or live weight (BW)] or carcass traits [e.g., hot standard carcass weight (HSCW) or carcass P8 rump fatty (CP8)] that are provided as outputs by the BeefSpecs calculator. The authors are unaware of any models for predicting ossification.

The objective of this study was to develop a prediction equation that uses either live animate being or carcass characteristics to predict ossification which would exist used as an input for phenotypic prediction of meat quality on alive animals.

MATERIALS AND METHODS

Animal intendance and use committee approving was not obtained for this study considering data were obtained from existing databases [New South Wales Department of Chief Industries (NSW DPI) and branch research centre (CRC) for Beef Genetic Technologies databases].

This study developed models that depict the relationships betwixt carcass ossification and either live animate being or carcass traits in beef cattle. Data used during model development were taken from a growth and slaughter experiment that used progeny born in 2012 from the NSW DPI muscling herd. Datasets used for model evaluation were taken from the NSW DPI muscling herd (built-in 2011) and the CRC for Cattle and Beef Quality (built-in 1993 to 1998) separated on the ground of Bos taurus or Bos indicus breed blazon (Robinson, 1995, Upton et al., 2001).

Data for Model Development

30-half-dozen Angus heifers and 58 Angus steers (94 animals in total) were used to develop ossification prediction models. Cafe et al. (2018) described the design and management of the NSW DPI muscling herd. In brief, the NSW DPI muscling herd was established in 1992 from an unselected Hereford herd that was mated to Angus bulls selected on visual muscle score (McKiernan, 2007). To create divergent muscle lines, the female person progeny and subsequent generations were also selected on muscle score and mated to Angus bulls purchased from inside the commercial Australian beef manufacture once again based on visual muscle score (Walmsley and McKiernan, 2011).

The 2012 born accomplice used to develop the models in this study were backgrounded following weaning on temperate pastures in North Eastern New South Wales. The animals were serially slaughtered during late 2013 and 2014 in 3 groups at boilerplate ages of 448, 532, and 626 d, respectively. The concluding slaughter group was grain finished for 97 d prior to slaughter.

All animals had BW (kg), visual muscle scores (McKiernan 2007), sex (heifer or steer), nativity, and measurement dates recorded. The birth and recording dates were used to calculate age at recording (AGEREC, days) and ADG (kg/d) between weaning and slaughter for each individual.

Ultrasound scanning was used to measure rib fat (USRIB, mm), subcutaneous rump fat (USP8, mm), and center musculus area (USEMA, cmⁱⁱ) on the live animal. These traits and the corresponding BW were recorded between iii and five times for each animate being, including immediately prior to slaughter, with individuals killed at older ages having more than opportunities for measurement. The USRIB and USEMA (M. longissimus thoracis et lumborum) measurements were taken at the twelfth and 13th ribs (Barwick et al., 2009), the USP8 measurement was taken at the intersection betwixt a line parallel to the spine from the tuber ischium and a line perpendicular to information technology from the barbed procedure of the third sacral vertebra (Johnston et al., 2003; Wolcott et al., 2009). All ultrasound measurements were taken by an accredited technician (Upton et al., 1999) with an Aloka 500V existent-time ultrasound scanner using a 17-cm transducer (Corometrics Medical Systems Inc., Wallingford, CT), with vegetable oil equally the coupling agent, and the fatty depths were recorded using callipers built into the scanner (Wolcott et al., 2001).

Following slaughter, carcass measurements were taken by MSA grading and AUS-MEAT chiller cess accredited graders. Ossification score, the degree of conversion of cartilage to bone, was measured at the sacral, lumbar, and thoracic vertebrae on a scale between 100 and 590 in increments of x (Wolcott et al., 2009). Hot standard carcass weight was measured in kilograms from untrimmed carcasses. Carcass EMA (CEMA) was measured using the AUS-MEAT EMA standard grid as the number of square centimetres (cmⁱⁱ) of longissimus thoracis et lumborum at the quartering site. Carcass rib fat (CRIB, mm) was measured equally the depth of subcutaneous fatty at the quartering site in the chilled carcass approximately 75% of the style forth the rib eye musculus. Carcass P8 rump fat was measured in millimeters as the depth of subcutaneous fatty at the intersection of a line parallel to the spine, from the tuber ischium, and a line perpendicular to it, from the spinous process of the third sacral vertebra (Johnston et al., 2003; Wolcott et al., 2009). MSA marbling score, a measure of the fat deposited between myofiber bundles in the rib eye muscle, was subjectively assessed using a scale between 100 and ane,100 in increments of x. Marbling was assessed at the quartering site of the chilled carcass a minimum of 45 min later on the carcass was quartered and is calculated by evaluating the amount, piece size, and distribution of marbling in comparison to the MSA reference standards (AUS-MEAT, 2005; Wolcott et al., 2009).

Animals with missing measurements were removed from the dataset. Summary statistics for the traits recorded on the live animal and carcass are shown in Table 1.

Table 1.

Summary statistics for the live animal and carcass traits recorded on Angus heifers and steers in the development dataset

	Live Animal						Carcass
Sexual practice	Traitone	due north	Min.	Max.	Avg.	SD	Trait2	due north	Min.	Max.	Avg.	SD
Heifer	AGEREC (d)	36	434	647	531.6	72.5	AGESL (d)	36	436	649	533.half-dozen	72.five
	BW (kg)	36	309	640	427.3	98.5	HSCW (kg)	36	146	332	215.8	56.8
	ADG (kg/d)	36	0.34	0.92	0.59	0.17	CRIB (mm)	36	one	12	6.1	3.5
	USRIB (mm)	36	2	14	7.1	iii.2	CP8 (mm)	36	1	27	11.4	7.0
	USP8 (mm)	36	3	20	x.6	4.9	Marble Score	36	200	510	329.ii	74.4
	USEMA (cm2)	36	50	84	64.6	9.9	CEMA (cmtwo)	36	44	99	66.1	xiii.iii
	Muscle Score	36	3	11	7.2	2.1	OSS	36	120	200	153.3	21.i
Steer	AGEREC (d)	58	420	648	531.4	75.ane	AGESL (d)	58	422	650	533.iv	75.i
	BW (kg)	58	327	740	493.9	111.5	HSCW (kg)	58	163	380	250.half dozen	63.v
	ADG (kg/d)	58	0.29	1.fifteen	0.seventy	0.xx	CRIB (mm)	58	i	13	5.0	3.4
	USRIB (mm)	58	2	14	half dozen.iii	iii.ii	CP8 (mm)	58	1	26	7.8	6.2
	USP8 (mm)	58	2	17	8.1	4.4	Marble Score	58	190	480	305.7	59.7
	USEMA (cmii)	58	52	92	69.v	11.3	CEMA (cm2)	58	43	110	69.iv	13.5
	Muscle Score	58	two	13	7.8	2.six	OSS	58	100	150	129.0	14.8

Data for Model Evaluation

Three datasets were used to evaluate the models developed and are described below. The traits were recorded in a like manner to that described for the development dataset. Summary statistics for OSS, AGEREC, BW, AGESL, HSCW, and CP8 in these datasets are given in Table 2.

Table two.

Summary statistics for historic period at measurement (AGEREC), live weight (BW), age at slaughter (AGESL), hot standard carcass weight (HSCW), carcass P8 fatty depth (CP8), and ossification (OSS) in each evaluation dataset

		Dataset 1					Dataset 2					Dataset 3
Sex	Trait	n	Min.	Max.	Avg.	SD	n	Min.	Max.	Avg.	SD	n	Min.	Max.	Avg.	SD
Heifer	Live
	AGEREC (d)	-	-	-	-	-	83	496	821	617.two	93.0	404	307	1131	729.vi	156.8
	BW (kg)	-	-	-	-	-	83	342	614	472.7	75.viii	404	172	678	434.seven	85.0
	Carcass
	AGESL (d)	-	-	-	-	-	83	501	838	625.5	95.7	404	429	1153	747.7	148.half-dozen
	HSCW (kg)	-	-	-	-	-	83	155	324	238.3	48.4	404	140	358	236.0	41.0
	CP8 (mm)	-	-	-	-	-	83	2	22	10.0	iv.eight	404	3	30	12.3	five.0
	OSS	-	-	-	-	-	83	140	200	169.8	16.viii	404	120	200	171.0	17.four
Steer	Live
	AGEREC (d)	74	519	690	609.ix	52.5	1202	340	961	630.three	154.ii	482	318	1290	756.5	173.1
	BW (kg)	74	391	736	538.1	87.3	1202	232	762	504.7	104.5	482	216	778	519.3	97.7
	Carcass
	AGESL (d)	74	544	693	623.9	42.2	1202	343	964	659.6	140.0	482	428	1297	768.3	169.6
	HSCW (kg)	74	226	438	312	57	1202	121	400	275.0	54.1	482	146	416	286.2	50.ix
	CP8 (mm)	74	4	26	12.ane	five.six	1202	ane	34	nine.four	4.half dozen	482	1	27	11.1	four.8
	OSS	74	100	150	133.six	10.viii	1202	100	200	143.4	20.5	482	100	200	148.9	19.viii

Dataset 1.

Information were obtained from an earlier cohort of the NSW DPI muscling herd as described above for the development dataset. This dataset independent 74 steers born in 2011 that were managed/recorded in a similar manner to the animals in the development dataset and slaughtered during early on 2013.

Dataset 2.

Data from Bos taurus breed types (Angus, Hereford, Shorthorn and Murray Gray) were collected within the CRC for Cattle and Beef Quality (Robinson, 1995; Upton et al., 2001). Data was nerveless between 1997 and 1999, and after removing animals (n = xix) with missing measurements or ossification scores outside of the range (OSS: between 100 and 200) contained in the development dataset, this dataset contained 83 heifers and 1,202 steers. Calves were purchased at weaning by the Beef CRC from 34 commercial breeder herds throughout eastern Australia and one-half were grown out on pasture and half in a feedlot (both occurred in North Eastern New South Wales). Grain finished cattle were targeted for feedlot entry weights of 300, 400 (short fed), and 400 kg (long fed) for 400, 520, and 640 kg slaughter BWs, respectively. Live animal traits were recorded at intervals of 6 mo and animals were slaughtered when the average BW of the cohort was predicted to achieve target carcass weights of 220 kg (domestic market), 280 kg (Korean market) or 340 kg (Japanese market place) (Upton et al., 2001). For more detail on the recording of slaughter traits, see Perry et al. (2001).

Dataset 3.

Data from B. indicus breed types (Brahman and Santa Gertrudis) was obtained from the Beef CRC as described for dataset 2. These calves were bred between 1995 and 1998 on 12 Queensland commercial properties and, after removing animals (northward = 73) with missing measurements or ossification scores exterior of the range (OSS: between 100 and 200) contained in the development dataset, this dataset contained 404 heifers and 482 steers. At weaning, one-3rd of the calves were transferred to join the animals in dataset 2 in Northern New S Wales and the rest was kept in Central Queensland for either pasture or feedlot finishing (Upton et al., 2001). These animals were slaughtered between 1997 and 2000. The same traits and measurement procedures were used for these animals as described for dataset ii.

Statistical Analysis

Linear regression analyses were conducted using the linear model process in the R statistical package (R Development Core Team, 2014) to develop prediction models with ossification as the dependent variable (Eq. [1]), as follows:

Ossification = β 0 + β 1 ten age / BW / HSCW + β 2 ten sexual activity + … .. + interactions + eastward ,

(ane)

where combinations of traits were fitted as covariates and fixed effects, with β_i representing the regression coefficient of each fitted covariate/fixed effect; and due east the residual error. All 2-way interactions were evaluated. Sex was coded equally heifer = 0 and steer = 1. The linear regressions were progressively developed, first with models that just contained age (AGEREC or AGESL), BW, or HSCW as covariates. The upshot of sex was then included in association with historic period, BW, and HSCW. Models that included either all live animal (USRIB, USP8, USEMA, muscle score and either BW or ADG) or carcass (CRIB, CP8, marbling score, CEMA and either HSCW or ADG) traits plus all 2-way interactions were as well evaluated. Simply one measure of growth was used in each of the models; thus, simply BW or ADG were included in each live trait model and only HSCW or ADG in the carcass trait models. Effects that were not significant (P < 0.05) were removed. The Box-Cox procedure (Box and Cox, 1964) in R was used to determine that the ossification information did non crave transformation. The adjusted R ⁱⁱ (R ² _adj) and remainder standard error (SE) were used to assess how well models fitted the development data. The R ² _adj (Eq. [2]) was used to adjust for the different number of parameters fitted in each model during model development:

R 2 adj = 1 − [ S S E ( n − one ) S Due south T ( n − p − 1 ) ] ,

(2)

where SST is the total sums of squares, SSE is the mistake sums of squares, due north is the number of observations, and p is the number of parameters in the model.

The prediction equations with R ^two _adj above 0.55 were challenged using the three evaluation datasets. The R statistical package (R Development Cadre Team, 2014) was used to plot observed versus predicted ossification with a line of unity (y = x) and the residuals (observed – predicted) versus predicted. The observed versus predicted plots with the line of unity do not have regression lines considering the accurateness of prediction is how close a dataset tin lie on the line of unity rather than the strength of an R ⁱⁱ value (Tedeschi 2006). Model evaluation was conducted using a customized process in the R statistical packet. Model predictions of ossification were evaluated using mean bias (MB; Eq. [3]):

where northward is the number of data points, O _i is the observed ossification, and P _i is the predicted ossification, respectively (i = one to due north). The error of prediction was assessed past the mean foursquare fault of prediction (MSEP; Eq. [4]):

where the terms are every bit defined in a higher place. The root mean square fault of prediction (RMSEP) was used as a measure out of the accuracy of prediction. The MSEP was decomposed into bias, gradient, and random components as a proportion of MSEP to appraise the error construction, following the method of Bibby and Toutenburg (1977). The statistical significance of each MB was evaluated using a paired t-test of the mean of the differences between the observed and model-predicted values.

RESULTS

Model Evolution

Comparing of models where either BW or ADG were included for the live animal traits and either HSCW or ADG were included for the carcass traits found in all instances that ADG was not significant in explaining variation in ossification and these models had lower fits to the ossification data than models containing either BW or HSCW. Consequently, just the results for models that contained BW or HSCW are presented.

The coefficients for the linear regressions of ossification on either alive animate being (sex, AGEREC, USRIB, USP8, USEMA, musculus score, grass or grain finishing, and BW) or carcass (sex, AGESL, CRIB, CP8, marbling score, CEMA, and HSCW) traits, including all 2-way interactions, are shown in Tabular array 3. The linear regression betwixt ossification and AGEREC (model A) demonstrates that AGEREC accounted for just a moderate proportion of the variation in ossification (R ⁱⁱ _adj = 0.30). Models C and E that contained either BW or HSCW only accounted for a very low proportion of the variation in ossification (both R ² _adj = 0.08). Models B, D, and F demonstrate that including sex activity and it's interaction with other traits deemed for more of the observed variation in ossification (P < 0.05). Models Chiliad and H demonstrate that CP8 fitted in conjunction with sex or sexual activity and HSCW also accounted for more of the observed variation in ossification (P < 0.05). Other live and carcass traits evaluated did not significantly (P > 0.05) business relationship for any of the variation observed in ossification. An unexpected outcome was that USP8 was not significant (P > 0.10) when tested during the development of model D, but CP8 was meaning in both models G and H.

Tabular array iii.

Linear regression coefficients (SE) for the prediction of ossification based on either live animal or carcass traits; adjusted R ² (R ⁱⁱ _adj) and residual standard error (SE) of model fit are presented

Model	Equation1	R2 adj	SE
Live Beast
A	54.28(13.33) + 0.16(0.02) × AGEREC	0.30	17.66
B	27.83(15.49) + 41.03(19.41)× sex + 0.23(0.03) × AGEREC – 0.12(0.04)× sex × AGEREC	0.66	12.38
C	111.79(9.09) + 0.06(0.02)× BW	0.08	twenty.23
D	85.35(10.16) + 0.16(0.02)× BW + 10.94(13.01)× sex – 0.09(0.03)× sex activity × BW	0.59	13.51
Carcass traits
E	115.10(8.eighteen) + 0.10(0.03)× HSCW	0.08	20.27
F	92.77(9.08) + 0.28(0.04)× HSCW + ten.85(xi.70)× sex – 0.18(0.05)× sexual practice × HSCW	0.58	13.70
G	126.53(4.29) – 6.06(5.15)× sex + ii.35(0.32)× CP8 – 1.25(0.43)× sexual practice × CP8	0.threescore	thirteen.35
H	107.15(ten.92) + xi.53(11.45)× sexual activity + 1.x(0.49)× CP8 + 0.xvi(0.07)× HSCW – 0.15(0.05)× sex × HSCW	0.60	thirteen.39

The R ^two _adj and SE shown in Table 3 signal models A, C, and E fit the ossification information less well than the other models. Model B, which contains sex, AGEREC, and their interaction, has the highest R ^two _adj and lowest SE (R ² _adj = 0.66, SE = 12.38). Models D, F, G, and H, which do not include AGEREC, but include BW, CP8, and/or HSCW, have slightly lower R ⁱⁱ _adj and slightly higher SE (R ² _adj = 0.59, 0.58, 0.60, and 0.60, SE = thirteen.51, 13.70, thirteen.35, and 13.39, respectively; Table 3) than model B.

Predictions of ossification are illustrated in Figs. 1 to ​ 3 for models B, D (alive inputs), F, M, and H (carcass inputs). The effect AGEREC, BW, HSCW, and CP8 have on the prediction of ossification is conspicuously demonstrated. Comparison of Fig. 2b and Fig. 3a illustrate the different influences CP8 has on the prediction of ossification, with divergent responses to CP8 in steers and heifers predicted by model G and parallel responses predicted by model H.

Predictions of ossification using (a) model B with age at scanning ranging from 436 to 649 d for heifers and 422 to 650 d for steers and using (b) model D with live weight ranging from 309 to 640 kg for heifers and 327 to 740 kg for steers.

Predictions of ossification using (a) model F with hot standard carcass weight ranging from 155 to 361 kg for heifers and 170 to 404 kg for steers and using (b) model Grand with carcass P8 fat (CP8) ranging from 1 to 27 mm in heifers and ane to 26 mm in steers.

Predictions of ossification using model H with (a) hot standard carcass weight held constant at 253 kg while carcass P8 fatty (CP8) ranged from 1 to 27 mm in heifers and from 1 to 26 mm in steers and with (b) carcass P8 fat (CP8) constant at ix mm while hot standard carcass weight ranged from 155 to 361 kg in heifers and from 170 to 404 kg in steers.

Model Evaluation

The 5 models with R ^two _adj above 0.55 in the development analyses were evaluated using the iii evaluation datasets with the results reported in Table 4. It is important to behave in heed when examining these evaluation results that ossification is assessed in increments of ten (Wolcott et al. 2009) every bit opposed to on a continuous scale. The MB in Tabular array 4 for dataset i indicates that all models predicted ossification accurately with model B having the largest bias (−4.2). Model D slightly under predicted ossification while models B, F, G, and H tended to slightly over predict. The MBs institute in dataset ii were higher than those found in dataset ane except model B which had an MB of two.8. Models D and G had the highest MB while F had the lowest, after model B. The MBs establish in dataset 3 were more often than not higher than those found in datasets 1 and 2 with model D having the highest MB and model F the lowest. In full general, all models tended to under predict ossification in datasets 2 and three, except model B in dataset 3, which over predicted ossification by a similar magnitude to the under prediction made past the other models.

Table four.

Evaluation of models B and D, that contain live traits, and models F, Thousand, and H, that incorporate carcass traits, for their accurateness of predicting ossification

Dataset:	Dataset 1					Dataset 2					Dataset 3
	Model					Model					Model
	Live Traits		Carcass Traits			Live Traits		Carcass Traits			Live Traits		Carcass Traits
Detail	B	D	F	Thou	H	B	D	F	G	H	B	D	F	One thousand	H
Mean observed	133.6	133.six	133.6	133.6	133.6	145.1	145.1	145.one	145.1	145.1	159.0	159.0	159.0	159.0	159.0
Hateful predicted	137.8	131.9	135.2	133.seven	134.1	142.three	131.7	133.3	132.0	132.five	175.3	141.5	144.six	143.0	144.i
Mean bias	−4.2	1.7	−one.half dozen	−0.ane	−0.5	2.viii	13.iv	11.8	13.1	12.half dozen	−16.3	17.5	fourteen.4	xvi.0	14.9
P-value	<0.01	0.130	0.175	0.984	0.700	<0.01	<0.01	<0.01	<0.01	<0.01	<0.01	<0.01	<0.01	<0.01	<0.01
MSEPone
RMSEP	ten.5	ten.0	9.viii	10.seven	10.vii	nineteen.6	23.0	22.1	23.seven	23.ii	31.9	25.7	23.iii	24.6	23.4
Bias, %	sixteen.0	iii.1	2.5	0.0	0.2	ii.1	34.2	28.9	30.9	29.half dozen	26.one	46.2	38.0	42.two	40.v
Slope, %	1.0	1.v	i.0	7.8	8.5	xvi.0	0.0	0.ane	0.ane	0.0	48.4	2.7	2.4	1.5	1.one
Random, %	83.0	95.four	96.5	92.2	91.iii	81.9	65.8	71.0	69.0	70.iv	25.5	51.1	59.6	56.3	58.four

The RMSEP followed a similar trend to that observed for MB across the three evaluation datasets (Tabular array 4). The average RMSEP was lowest in dataset 1, increased in dataset ii and increased again in dataset 3. The RMSEP were similar in dataset 1 for all models. In dataset two model B had the lowest RMSEP (19.6), while it had the highest (31.9) in dataset three. The remaining models had similar RMSEP in both datasets two and three. The decomposition of the MSEP into its components indicated that the majority (>fifty%) of fault was of a random nature, except model B in dataset 3 where the random component only accounted for ~25% of the error. The proportion of random fault on average decreased from dataset one to dataset 2, and finally to dataset 3. In dataset 1, model B had some error contained in the bias component while the remaining models had error contained in the slope component. In datasets 2 and iii, the opposite occurred, where model B had more error contained in the slope component and the remaining models had more error in the bias component. The predictive power of model H in each dataset across the range of ossification scores present in the development dataset is shown in Fig. iv.

Evaluation of model H for datasets 1, ii, and 3 showing plots of (a) observed versus predicted and (b) the residue (observed – predicted) for ossification.

DISCUSION

The cardinal to remaining profitable and producing meat sustainably in any product organisation is to produce the product the consumer demands consistently and toll effectively (Egan et al., 2001; McPhee et al., 2006). Phenotypic prediction systems (e.g., BeefSpecs calculator; Walmsley et al., 2014), have been adult to provide cattle producers with data that would otherwise not exist attainable to assist inform management decisions that touch carcass compliance and meat quality. The increasing focus on the consumer eating experience necessitates that these prediction systems broaden their capabilities to include non only compliance traits similar HSCW and CP8 but meat quality traits.

Tenderness is well established as being disquisitional in determining meat eating quality and acceptability (Voges et al., 2007; Destefanis et al., 2008) with consumer preference studies of sensory attributes showing tenderness is more important than flavor or juiciness (Tornberg 1996; Destefanis et al., 2008). This is further demonstrated by the positive human relationship between the price of a cut of meat and its tenderness (Miller et al., 2001). The negative relationship betwixt tenderness and brute maturity (Romans et al., 1965; Shorthose and Harris 1990; Schonfeldt and Strydom 2011) due to muscle tissue containing decreasing amounts of estrus-liable collagen every bit animals mature (Schonfeldt and Strydom 2011) has been seized upon past meat grading systems (e.g., MSA, USDA) as a valuable indicator of eating quality and consumer acceptability.

Previous research has used different measures as indicators of animal maturity including chronological age (Shackelford et al., 1995; Raines et al., 2008) and ossification (Romans et al., 1965; Shackelford et al., 1995; Raines et al., 2008). Although these measures take been shown to be positively correlated (0.64, Raines et al., 2008), they are not synonymous (Romans et al., 1965). Age is a linear measure that is not affected by physiological processes such every bit pregnancy, lactation, or age at maturity, whereas ossification is strongly influenced by hormonal status of the animal (Field et al., 1997; Scheffler et al., 2003). This close association with physiological processes allows ossification to improve reflect physiological maturity (Field et al., 1997), and thus tenderness (Weston et al., 2002) than age and is why the MSA and USDA grading systems both use ossification as the measure for maturity when predicting eating quality.

Model Development

The linear regression betwixt ossification and AGEREC (model A) supports that they are non synonymous in that AGEREC accounted for only a moderate proportion of the variation nowadays in ossification (R ² _adj = 0.xxx, Table 3). Despite AGEREC explaining less than anticipated information technology explained a significantly larger proportion of the variation in ossification than either BW (model C) or HSCW alone (model E; both R ² _adj = 0.08, Table three). Although looking at age, Raines et al., (2008) found a correlation with hot carcass weight of 0.12 which contrasts to the correlation with ossification of 0.64 which suggests that weight in isolation is not a good indicator of historic period or ossification. Regressing ossification on ADG (results not presented) provided no advantage over regressing on BW or HSCW which was again unexpected given animals of a similar type with college ADG and thus younger ages at a detail weight would exist expected to be less physiologically mature. This result is supported by Field et al. (1997) who plant that 2 similar groups of heifers had significantly different (P < 0.01) ossification, as shown past bone maturity, after similar rates of weight gain.

The charge per unit of development of ossification and hence the ossification score of a carcass is strongly influenced by hormonal status of the animal, and in particular oestrogen (Field et al., 1997; Scheffler et al., 2003). Oestrogen levels at any signal of development of the animal are influenced by sexual practice, castration, hormonal growth promotants, and parity status (Waggoner et al., 1990; Field et al., 1996; Scheffler et al., 2003). The variation in ossification explained when sex was included with AGEREC (model B) doubled compared to AGEREC solitary (model A) whereas when sex was included with either BW (model D) or HSCW (model F) the variation in ossification explained increased by more than than 7 times (Table 3) compared to either BW (model C) or HSCW (model Due east) solitary. When sex activity and CP8 were included, slightly more variation (R ^two _adj = 0.60) was explained than when sexual activity was included with either BW (R ² _adj = 0.59) or HSCW (R ⁱⁱ _adj = 0.58). The design of accession of body components where fat accumulation takes higher priority than os or lean tissue every bit animal'southward mature (Berg and Butterfield 1976) provides a possible caption as to why CP8 when used in clan with sex explained slightly more variation in ossification than either BW or HSCW. The inclusion of HSCW with sex and CP8 explained the aforementioned corporeality of variation in ossification (R ² _adj = 0.60) equally sex and CP8 fat which could be interpreted every bit weight may have a limited capacity to draw physiological maturity. However, this was not observed during the development of model D where USP8 was not pregnant (P > 0.10) in explaining the variation in ossification while BW was meaning (P < 0.05).

Model Evaluation

The purpose of this inquiry was to develop a prediction equation that uses either live animal or derived carcass traits to predict ossification with sufficient accuracy for phenotypic prediction systems to make robust predictions of meat quality. Model evaluation revealed that, in general, all the models tested had similar accuracies within a dataset and like differences in accuracies between datasets. The merely exception was model B which had lower absolute MB and RMSEP when tested in dataset 2 than the other models and a higher RMSEP than the other models in dataset 3. In reality, these differences are not as big as they initially appear due to the calibration (increments of 10) used for assessing ossification (Wolcott et al. 2009). The RMSEP for all models in all evaluation datasets was within, or shut to, 3 increments of the ossification calibration while the MB was inside 2 increments. In particular, the MB in dataset i was substantially nil and the RMSEP was ane when the ossification scale is taken into consideration. These results signal all models have predictive accurateness sufficient for inclusion in phenotypic prediction systems.

When examining the predictive accurateness of models between datasets all models generally had college predictive accuracy in dataset 1 than dataset ii and lower accurateness again in dataset 3. The loss in accuracy for all models except model B when tested in datasets 2 and 3 compared to dataset 1 is attributable to the bias component of the RMSEP. This event suggests that there is a factor(s) impacting on ossification that differs between dataset ane and datasets 2 and 3 which are non being accounted for by any of the models. This could be explained by dataset 1 beingness an earlier cohort of the aforementioned herd from which the development dataset was obtained whereas datasets 2 and 3 are regarded as being unrelated to the evolution dataset. Model B had more than mistake in the gradient component than the bias component of RMSEP which suggests that there is a unlike human relationship between ossification and AGEREC in datasets 2 and 3 than the evolution dataset or dataset 1. This again could be explained past dataset ane being recorded in the 12 mo prior to the evolution dataset with the ii datasets experiencing a similar production environment and like genetic backgrounds. In contrast, datasets 2 and 3 were recorded approximately 15 yr earlier and both the genetics of the animals and production systems take changed.

The evolution of the method used to record ossification score in Commonwealth of australia could too provide some explanation of the evaluation differences seen between dataset 1 and datasets 2 and 3. The ossification data in datasets 2 and iii was actually recorded earlier ossification was adopted by MSA (Perry et al., 2001) for describing the affect of animal maturity on meat quality. Information technology is feasible that developments which take occurred during the development of MSA and the grooming of the MSA technicians could take increased the quality of trait recording constitute in dataset one and the evolution dataset compared to datasets ii and 3. This may also explicate why unpublished results described below (using datasets 2 and 3 to develop models) produced less accurate results than those reported in this study.

The decrease in accuracy found between datasets ii and three is explained by the differences in brood composition between the evolution dataset and the evaluation datasets. Dataset 1 was an earlier cohort of the same herd from which the development dataset was obtained; dataset 2 was unrelated but comprised, predominantly, of the same breeds as found in the development dataset, while dataset 3 was comprised of B. indicus content animals. The correlation betwixt the accuracy with which the models predicted ossification and the relatedness of the validation datasets in terms of breed content suggests that the prediction model(s) may require a breed cistron to be included. The absence of variation in breed in the development dataset prevented this from occurring in this study.

In an attempt to overcome this limitation, a combination of datasets 2 and 3 were used to develop prediction models that included a breed factor but this proved to be unsuccessful (unpublished). The prediction models obtained when doing this had higher MB and a greater proportion of the MSEP due to slope and bias when evaluated using dataset 1 and the development dataset. These MB and MSEP values were all larger than those values presented in Table 4 for the models in the current study. An alternative would be to include an ad hoc B. indicus breed correction, based on the MB found in dataset three evaluations, in the models to increment their transferability to other breeds. Even so, this would require a second B. indicus dataset to evaluate this approach, which was not available in this study. It would be highly desirable to include variation in breed content in whatsoever futurity research focused on ossification.

Model Selection for Phenotypic Prediction

The like predictive accuracy observed for all models evaluated across the 3 evaluation datasets (Table 4) when because ossification is recorded in 10 unit increments (Wolcott et al., 2009) indicates that any of the models could exist used to make phenotypic predictions of ossification. The context in which the equations are used will ultimately make up one's mind their usefulness. Given very few cattle in commercial production systems accept birth dates recorded and thus age known, model B is anticipated to be of limited value for predicting ossification. It is desirable to minimize the number of input parameters to those necessary for accurate prediction, however models D, F, M, and H comprise but 2 or 3 input parameters and one of these is sex activity. If ossification was to be predicted contained of any other modelling system, model D would probably be the well-nigh suitable given the inputs (BW and sex) are more often than not accurately recorded. The potential weakness of model D is that 2 animals of equal BW, but unlike mature sizes would be predicted to have the same ossification even though the creature with a lower mature size would be expected to have a college physiological maturity (Berg and Butterfield 1976), and thus ossification (Romans et al., 1965; Shackelford et al., 1995; Raines et al., 2008) at whatever weight.

When fat deposition increases in priority relative to other tissues equally animal's mature, animals with lower mature size would be expected to be fatter at any given BW/HSCW which is captured in model H. For both sexes at any HSCW, model H would predict ossification to be higher every bit CP8 fatty increases (Fig. 3). This is not represented in model G, in fact model Chiliad suffers a similar weakness to model D, in that, animals which grew at 2 different rates to the aforementioned final CP8 fatty depth would exist predicted to have the same ossification even though the beast which grew at the slower rate would exist expected to take a higher physiological maturity, and thus ossification. The weakness of model H is HSCW and fifty-fifty CP8 fat depth are not elementary measures to record accurately on live animals. However, when used in association with a phenotypic prediction system like the BeefSpecs calculator (Walmsley et al., 2014), which predicts carcass P8 rump fatty and HSCW, model H could produce predictions of ossification that align with physiological expectations.

The accuracy of the predictions of ossification would be determined by the accuracy of the predictions of P8 rump fat and HSCW. Only when HSCW is predicted with higher accurateness than the accurateness of measuring animal live weight would the accuracy of models E, F, and H exist expected to exist higher than model D when assessing animals ready for slaughter. When predicting into the time to come, every bit is the intent of the BeefSpecs calculator, predictions of ossification made with model D would exist limited by the accuracy of predicting hereafter live weight. Using whatsoever of the models developed and evaluated in this study in the appropriate context to inform predictors of meat quality volition assist beef producers focus more of their determination making on satisfying consumer demands, as described past carcass grading systems, consistently and cost effectively to remain profitable. The collection of information that includes variation in breed is desirable to further develop the models in this report to increase their accuracy of prediction.

Footnotes

The authors acknowledge all participants, both scientists and technical staff, who contributed to or supported the piece of work, including those involved in experimental blueprint, cattle management, data collection, and data handling. Funding for this study was supplied by Meat and Livestock Australia.

LITERATURE CITED

• AUS-MEAT 2005. Handbook of Australian meat. seventh ed. AUS-MEAT Ltd, Brisbane, QLD. [Google Scholar]
• Barwick S. A., Wolcott M. Fifty., Johnston D. J., Burrow H. M., and Sullivan M. T.. 2009. Genetics of steer daily and residual feed intake in two tropical beef genotypes, and relationships amongst intake, body limerick, growth and other post-weaning measures. Anim. Prod. Sci. 49:351–366. doi: ten.1071/EA08249 [CrossRef] [Google Scholar]
• Berg R. T., and Butterfield R. M.. 1976. New concepts of cattle growth. Sydney University press, Sydney, NSW. [Google Scholar]
• Bibby J., and Toutenburg H.. 1977. Prediction and improved estimation in linear models. John Wiley & Sons and Akademie-Verlag, Berlin, Gdr. [Google Scholar]
• Box M. Due east. P., and Cox D. R.. 1964. An analysis of transformations (with word). J. R. Stat. Soc. B. Stat. Methodol. 26:211–252. http://www.jstor.org/stable/2984418. [Google Scholar]
• Buffet L. M., McKiernan W. A., and Robinson D. 50.. 2018. Selection for increased muscling is not detrimental to maternal productivity traits in Angus cows. Anim. Prod. Sci. 58:185–192. doi: 10.1071/AN13023 [CrossRef] [Google Scholar]
• Destefanis G., A. Brugiapaglia M. T. Clomp, and Dal Molin E.. 2008. Human relationship between beef consumer tenderness perception and Warner-Bratzler shear forcefulness. Meat Sci. 78:153–156. doi: 10.1016/j.meatsci.2007.05.031 [PubMed] [CrossRef] [Google Scholar]
• Egan A. F., Ferguson D. M., and Thompson J. G.. 2001. Consumer sensory requirements for beefiness and their implications for the Australian beef industry. Aust. J. Exp. Agric. 41:855–859. doi: 10.1071/EA00065 [CrossRef] [Google Scholar]
• Field R., R. McCormick V. Balasubramanian D. Sanson J. Wise D. Hixon M. Riley, and Russell W.. 1996. Growth, carcass, and tenderness characteristics of virgin, spayed, and single-dogie heifers. J. Anim. Sci. 74:2178–2186. doi:10.2527/1996.7492178x [PubMed] [Google Scholar]
• Field R., R. McCormick Five. Balasubramanian D. Sanson J. Wise D. Hixon Yard. Riley, and Russell W.. 1997. Tenderness variation amidst loin steaks from A and C maturity carcasses of heifers similar in chronological age. J. Anim. Sci. 75:693–699. doi: x.2527/1997.753693x [PubMed] [CrossRef] [Google Scholar]
• Johnston D. J., Reverter A., Burrow H. M., Oddy V. H., and Robinson D. L.. 2003. Genetic and phenotypic characterisation of beast, carcass, and meat quality traits from temperate and tropically adapted beef breeds. 1. Animal measures. Aust. J. Agric. Res. 54:107–118. doi: 10.1071/AR02085 [CrossRef] [Google Scholar]
• McKiernan Due west. A. 2007. Muscle scoring beef cattle. Primefact No. 328. NSW Department of Master Industries, Orange, NSW. [Google Scholar]
• McPhee M. J., J. West. Oltjen T. R. Famula, and Sainz R. D.. 2006. Meta-analysis of factors affecting carcass characteristics of feedlot steers. J. Anim. Sci. 84:3143–3154. doi: 10.2527/jas.2006-175 [PubMed] [CrossRef] [Google Scholar]
• Miller M. F., G. A. Carr C. B. Ramsey K. 50. Crockett, and Hoover Fifty. C.. 2001. Consumer thresholds for establishing the value of beef tenderness. J. Anim. Sci. 79:3062–3068. doi: x.2527/2001.79123062x [PubMed] [CrossRef] [Google Scholar]
• Perry D., Shorthose W. R., Ferguson D. M., and Thompson J. M.. 2001. Methods used in the CRC program for the determination of carcass yield and beef quality. Aust. J. Exp. Agric. 41:953–957. doi: x.1071/EA00092 [CrossRef] [Google Scholar]
• R Development Core Team 2014. R: a language and environment for statistical computing. R Foundation for Statistical Computing. R Foundation for Statistical Calculating, Vienna, Austria. [Google Scholar]
• Raines C. R., One thousand. E. Dikeman J. A. Unruh M. C. Hunt, and Knock R. C.. 2008. Predicting cattle age from eye lens weight and nitrogen content, dentition, and United States department of agriculture maturity score. J. Anim. Sci. 86:3557–3567. doi: 10.2527/jas.2007-0445 [PubMed] [CrossRef] [Google Scholar]
• Robinson D. L. 1995. Pattern of the CRC straightbred genetics experiments. Proc. Assoc. Advmt. Anim. Breed. Genet. 11:541–545. [Google Scholar]
• Romans J. R., H. J. Tuma, and Tucker W. 50.. 1965. Influence of carcass maturity and marbling on the physical and chemical characteristics of beefiness. I. Palatability, fiber bore and proximate assay. J. Anim. Sci. 24:681–685. doi: 10.2527/jas1965.243681x [PubMed] [CrossRef] [Google Scholar]
• Scheffler J. Yard., D. D. Buskirk S. R. Rust J. D. Cowley, and Doumit K. E.. 2003. Outcome of repeated administration of combination trenbolone acetate and estradiol implants on growth, carcass traits, and beef quality of long-fed holstein steers. J. Anim. Sci. 81:2395–2400. doi: x.2527/2003.81102395x [PubMed] [CrossRef] [Google Scholar]
• Schönfeldt H. C. and Strydom P. E.. 2011. Effect of historic period and cut on tenderness of Due south African beef. Meat Sci. 87:206–218. doi: 10.1016/j.meatsci.2010.10.011 [PubMed] [CrossRef] [Google Scholar]
• Shackelford S. D., M. Koohmaraie, and Wheeler T. 50.. 1995. Effects of slaughter age on meat tenderness and USDA carcass maturity scores of beef females. J. Anim. Sci. 73:3304–3309. doi: x.2527/1995.73113304x [PubMed] [CrossRef] [Google Scholar]
• Shorthose W. R. and Harris P. V.. 1990. Outcome of creature age on the tenderness of selected beef muscles. J. Food Sci. 55:1–8. doi: 10.1111/j.1365-2621.1990.tb06004.ten [CrossRef] [Google Scholar]
• Tedeschi Fifty. O. 2006. Assessment of the adequacy of mathematical models. Agric. Syst. 89:225–247. doi: x.1016/j.agsy.2005.xi.004 [CrossRef] [Google Scholar]
• Tornberg E. 1996. Biophysical aspects of meat tenderness. Meat Sci. 43:S175–S191. doi: ten.1016/0309-1740(96)00064-2 [PubMed] [CrossRef] [Google Scholar]
• Upton Westward. H., Couch H. Chiliad., Dundon A., Robinson D. L., and Farrell E. B.. 2001. CRC breeding plan design, measurements and database: methods that underpin CRC research results. Aust. J. Exp. Agric. 41:943–952. doi: 10.1071/EA00064 [CrossRef] [Google Scholar]
• Upton W. H., Donoghue Thousand. A., Graser H.-U., and Johnston D. J.. 1999. Ultrasound proficiency testing. Proc. Assoc. Advmt. Anim. Breed. Genet. 13:341–344. [Google Scholar]
• USDA 1997. United States standards for grades of slaughter cattle. Agricultural Marketing Service, USDA, Washington, DC. [Google Scholar]
• Voges One thousand. L., C. L., Mason J. C., Brooks R. J., Delmore D. B., Griffin D. S., Hale W. R., Henning D. D., Johnson C. 50., Lorenzen R. J., Maddock, et al. 2007. National beef tenderness survey - 2006: assessment of Warner-Bratzler shear and sensory panel ratings for beef from US retail and foodservice establishments. Meat Sci. 77:357–364. doi: 10.1016/j.meatsci.2007.03.024 [PubMed] [CrossRef] [Google Scholar]
• Waggoner A. W., M. Due east. Dikeman J. R. Brethour, and Kemp K. E.. 1990. Performance, carcass, cartilage calcium, sensory and collagen traits of longissimus muscles of open versus 30-calendar month-old heifers that produced one dogie. J. Anim. Sci. 68:2380–2386. doi: ten.2527/1990.6882380x [PubMed] [CrossRef] [Google Scholar]
• Walmsley B. J., and McKiernan West. A.. 2011. Observed trends in estimated breeding values in response to pick using visual muscle score in beefiness cattle. Proc. Assoc. Advmt. Anim. Breed. Genet. 19:123–126. [Google Scholar]
• Walmsley B. J., McPhee G. J., and Oddy 5. H.. 2014. Development of the BeefSpecs fat figurer to assist determination making to increase compliance rates with beef carcass specifications. Anim. Prod. Sci. 54:2003–2010. doi: 10.1071/AN14611 [CrossRef] [Google Scholar]
• Watson R., Polkinghorne R., and Thompson J. M.. 2008. Evolution of the Meat Standards Australia (MSA) prediction model for beef palatability. Aust. J. Exp. Ag. 48:1368–1379. doi: 10.1071/EA07184 [CrossRef] [Google Scholar]
• Weston A. R., Rogers R. W., and Althen T. G.. 2002. Review: the role of collagen in meat tenderness. Prof. Anim. Sci. 18:107–111. doi: x.15232/S1080-7446(15)31497-two [CrossRef] [Google Scholar]
• Wolcott M. L., Johnston D. J., Barwick Due south. A., Iker C. L., Thompson J. M., and Couch H. M.. 2009. Genetics of meat quality and carcass traits and the affect of tenderstretching in two tropical beef genotypes. Anim. Prod. Sci. 49:383–398. doi: ten.1071/EA08275 [CrossRef] [Google Scholar]
• Wolcott Thou. L., Thompson J. Thou., and Perry D.. 2001. The prediction of retail beef yield from real time ultrasound measurements on live animals at three stages through growout and finishing. Aust. J. Exp. Agric. 41:1005–1011. doi: ten.1071/EA00017 [CrossRef] [Google Scholar]

---

Articles from Journal of Animal Science are provided hither courtesy of Oxford University Press

---

rogerssheire.blogspot.com

Source: https://www.ncbi.nlm.nih.gov/pmc/articles/PMC6313154/

Share this post