*2.1. Animals*

A total of 111 cull cow left carcass sides (hot carcass weight: HCW = 346 ± 33.3 kg), sourced from a commercial abattoir (*n* = 72) and from the AAFC-Lacombe Research and Development Centre (AAFC-Lacombe RDC) cow herd (*n* = 39), were used in the present study. AAFC-Lacombe RDC animals were cared for according to the Canadian Council on Animal Care Guidelines [19] (AAFC-Lacombe RDC study plan No. 201705).

#### *2.2. Carcass Sides, Cut-Out, and CVS and DXA Scanning*

Cull cows sampled from the AAFC-Lacombe RDC herd were slaughtered at the AAFC-Lacombe RDC federally inspected abattoir. Following slaughter, carcasses were dressed and split and HCW were recorded. In turn, commercial carcass sides, harvested following the Guidelines for the humane care and handling of food animals at slaughter (Canadian Food Inspection Agency, CFIA) [20], were shipped to AAFC-Lacombe RDC facilities in a

refrigerated truck following guidelines for transportation of carcasses over thirty months of age (CFIA) [21]. At the time of slaughter, HCW were recorded by the personnel of the slaughter plant. In both sample populations, pictures of each carcass side were taken using a HCC unit VBS 2000, *e+v*® Technology GmbH, Oranienburg, Germany.

Raw output data of the HCC images were composed of 187 variables describing carcass dimensions: angle (W00-W99), length (L00-L19), area (F00-F13, F20-F29), carcass contour and volumes (V00-V13, V20-V29), and color (Fe00-Fe18). Following 72 h of chilling at 2 ◦C, left carcass sides were weighed (CCW) to determine shrink loss. Physiological maturity of the carcasses were assessed based on the extent to which caps of the spinal processes and ribs had ossified (i.e., >50% ossification, a carcass receives a D grade) in accordance with <sup>L</sup>ópez-Campos et al. [22] and the Canadian Beef Grading Agency (CBGA) [2]. Left carcass sides were then knife-ribbed between the 12th and 13th ribs. After 20 min of atmospheric oxygen exposure, full Canadian grading data were collected by a certified grader from the CBGA. The assessments included grade fat (minimum fat thickness over the rib in 4th quadrant from the spinous process, mm), fat thickness (at the three-quarters position from the spinous process, mm), rib-eye area (REA; in cm<sup>2</sup> of the *Longissimus thoracis*), and marbling score, subjectively assessed using United States Department of Agriculture (USDA) beef marbling pictorial standards as reference points [23]. Muscle scores (1–4) were also determined based on *L. thoracis* length and width, measured at the grade site [24,25]. Then, rib-eye pictures from each carcass side were taken using CCC: VBG 2000 (*e+v*® Technology GmbH, Oranienburg, Germany). Each image was then processed by manufacturer software, in real-time, to produce raw output data composed of 99 variables describing a number of measurements related to measurements on the rib-eye (*n* = 22), fat thickness (*n* = 15), and muscle and fat color (*n* = 15), as well as other variables (*n* = 47, e.g., marbling assessment, back fat dressing corrections, etc.).

Estimated total lean meat yield (LMY) was calculated according to the Jones et al. [24,25] equation LMY (%) = 63.5 + 1.05 × (muscle score) − 0.76 × (grade fat). The retail cut yield (RCY) percentage was calculated using the equation RCY (%) = 51.34 − 5.78 × (fat thickness at the 3 4 , inches) − 0.46 × (kidney, pelvic, and heart fat percent, KPH, %) − 0.0093 × (hot carcass weight, HCW, pounds) + 0.74 × (REA, square inches) [25].

Left carcass sides were fabricated into primal cuts with carcass breakpoints identified following the Institutional Meat Purchase Specifications (IMPS) for Fresh Beef Products, Series 100 [26]. The primals collected from the left fabricated carcass side were the chuck (IMPS #113), rib (IMPS #103), brisket (IMPS #118), flank (IMPS #193, non-trimmed), foreshank (IMPS #117), loin (IMPS #172A), round (IMPS #158A), and plate (IMPS #121). Following procedures described by <sup>L</sup>ópez-Campos et al. [18], each primal cut was scanned with a GE Lunar iDXA unit (GE Lunar, General Electric, Madison, WI, USA) using the whole-body scan option on standard mode to estimate fat, lean, and bone weights. After DXA scanning, all left primals were fully dissected into subcutaneous fat (SQ), intermuscular fat (IM), body cavity fat (BC), lean, and bone, then weighed by trained personnel. An adequate dissection processing was carried out by highly skilled meat cutters ensuring that the difference between primal weight and the sum of total bone, total lean, and total fat was not higher than a 2%.

### *2.3. Statistical Analyses*

All the statistical analyses were performed using SAS v. 9.4 (SAS Institute Inc., Cary, NC, USA, 2014) [27]. Either CVS or DXA estimates of lean, fat, and bone weights from each primal cut, and the overall fat, lean, and bone weights were included as independent variables in a partial least square regression (PLSR) to generate prediction equations. Therefore, four different groups were defined depending on the regression estimating variables used: HCC, CCC, combination of HCC + CCC, and DXA. All models were used to predict the reference values from the manual dissections and the calculations of LMY and RCY equations.

All PLSR models were fit using an internal full leave-one-out cross-validation, to avoid overfitting in the calibration set, and the number of latent variables (LV) used to minimize predicted residual error sums of squares (PRESS) was reported for the calibrated PLSR models.

The predictive ability of the PLSR models was evaluated in terms of coefficient of determination (R2) and the mean square prediction error (MSPE), which was decomposed into error in central tendency (ECT), error due to regression (ER), and error due to disturbances (ED) [18]. These three fractions were calculated and expressed as percentages, as suggested by Benchaar et al. [28], as a means of describing the residual error in the models. ECT indicates how the average of CVS/DXA values deviates from the average of dissection values. ER measures the deviation of the least square regression coefficient from one, which is the value that it would have been if dissection and CVS/DXA measurements were in complete agreement. The ED is the variation in dissection measurements that is not accounted for by the least square regression of CVS/DXA measurements. In fact, this error is the unexplained variance and represents the portion of MSPE that cannot be eliminated by linear correction of the predictions [29]. Finally, when expressed as a percentage of the MSPE, the ECT, ER, and ED are called bias proportion, regression proportion (deviation of the regression slope from one), and disturbance proportion, respectively [30].
