Application of biological simulation models in estimating feed efficiency of finishing steers

Data on individual daily feed intake, BW at 28-d intervals, and carcass composition were obtained on 1,212 crossbred steers. Within-animal regressions of cumulative feed intake and BW on linear and quadratic days on feed were used to quantify initial and ending BW, average daily observed feed intake (OFI), and ADG over a 120-d finishing period. Feed intake was predicted (PFI) with 3 biological simulation models (BSM): a) Decision Evaluator for the Cattle Industry, b) Cornell Value Discovery System, and c) NRC update 2000, using observed growth and carcass data as input. Residual feed intake (RFI) was estimated using OFI (RFI(EL)) in a linear statistical model (LSM), and feed conversion ratio (FCR) was estimated as OFI/ADG (FCR(E)). Output from the BSM was used to estimate RFI by using PFI in place of OFI with the same LSM, and FCR was estimated as PFI/ADG. These estimates were evaluated against RFI(EL) and FCR(E). In a second analysis, estimates of RFI were obtained for the 3 BSM as the difference between OFI and PFI, and these estimates were evaluated against RFI(EL). The residual variation was extremely small when PFI was used in the LSM to estimate RFI, and this was mainly due to the fact that the same input variables (initial BW, days on feed, and ADG) were used in the BSM and LSM. Hence, the use of PFI obtained with BSM as a replacement for OFI in a LSM to characterize individual animals for RFI was not feasible. This conclusion was also supported by weak correlations (<0.4) between RFI(EL) and RFI obtained with PFI in the LSM, and very weak correlations (<0.13) between RFI(EL) and FCR obtained with PFI. In the second analysis, correlations (>0.89) for RFI(EL) with the other RFI estimates suggest little difference between RFI(EL) and any of these RFI estimates. In addition, results suggest that the RFI estimates calculated with PFI would be better able to identify animals with low OFI and small ADG as inefficient compared with RFI(EL). These results may be due to the fact that computer models predict performance on an individual-animal basis in contrast to a LSM, which estimates a fixed relationship for all animals; hence, the BSM may provide RFI estimates that are closer to the true biological efficiency of animals. In addition, BSM may facilitate comparisons across different data sets and provide more accurate estimates of efficiency in small data sets where errors would be greater with a LSM.