Abstract Monitoring technologies allow to record phenotypic traits such as body weight (BW) and feed intake at very high frequencies. These traits can be analyzed using time-series analyses and there has been a long tradition to model BW as a function of time, using functions such as the logistic and Gompertz functions. Body weight can also be modeled as a function of cumulative feed intake (CFI), for example by the mono-molecular function. The first-derivative of this function describes the variation in the gain-to-feed ratio (and thus the nutrient requirements) during the growing period. The objective of this study is to describe a data analysis method that uses both approaches so that the variation in growth and feed intake patterns of growing pigs can be summarized in a limited number of model parameters. Body weight and CFI data of 92 growing pigs were recorded during approximately 80 days using an automated scale that weighed the pigs when they entered the area with automatic feeding stations. Data were recorded continuously and were analyzed per day. The data (i.e., time, BW, and CFI) were analyzed using a joint non-linear regression (i.e., one outcome (BW) with two regression equations) using the mono-molecular function to describe BW as a function of CFI (i.e., BW = (a1-c1)·(1exp(b1·CFI))+c1), and the Gompertz function to describe BW as a function of time (i.e., BW = a2·exp(-b2·exp(-c2·time))). Although each function relies on three parameters, they share the initial BW (i.e., when time=0 and CFI=0). The asymptotes of both models (i.e., a1 and a2) should also be identical but these are beyond the range of observations. The model was parameterized to include the initial BW, the time and feed required to gain 80 kg of BW (quantified as BW gain and feed conversion ratio (FCR)), and the shape parameters b1 and c2. Parameter estimates for the 92 pigs were 35.5 ± 4.4 kg for the initial BW, 2.66 ± 0.26 for the FCR, 0.99 ± 0.10 kg/day for BW gain, 0.0019 ± 0.0014 /kg feed for b1, and 0.0095 ± 0.0047 /day for c2. The correlations between the model parameters were low to moderate, ranging from 0.01 between the initial BW and FCR to 0.68 between the shape parameters of the two curves, indicating that each of the parameters provided unique phenotypic information about the animal. The results illustrate that variation among pigs is not only characterized by average performance traits (i.e., BW gain and FCR) but also by how these traits change during growth. The joint analysis of BW as functions of time and CFI can be used to phenotype animals for genetic selection and for precision livestock feeding.
Milgen et al. (Wed,) studied this question.