Bayesian simultaneous equation models for the analysis of energy intake and partitioning in growing pigs/ created by Anders Bjerring Strathe, Henry Jørgensen, E Kebreab and Allan Christian Danfær

The objective of the current study was to develop Bayesian simultaneous equation models for modelling energy intake and partitioning in growing pigs. A key feature of the Bayesian approach is that parameters are assigned prior distributions, which may reflect the current state of nature. In the models, rates of metabolizable energy (ME) intake, protein deposition (PD) and lipid deposition (LD) were treated as dependent variables accounting for residuals being correlated. Two complementary equation systems were used to model ME intake (MEI), PD and LD. Informative priors were developed, reflecting current knowledge about metabolic scaling and partial efficiencies of PD and LD rates, whereas flat non-informative priors were used for the reminder of the parameters. The experimental data analysed originate from a balance and respiration trial with 17 cross-bred pigs of three genders (barrows, boars and gilts) selected on the basis of similar birth weight. The pigs were fed four diets based on barley, wheat and soybean meal supplemented with crystalline amino acids to meet or exceed Danish nutrient requirement standards. Nutrient balances and gas exchanges were measured at c. 25, 75, 120 and 150 kg body weight (BW) using metabolic cages and open circuit respiration chambers. A total of 56 measurements were performed. The sensitivity analysis showed that only the maintenance component was sensitive to the prior specification, and hence the maintenance estimate of 0·91 MJ ME/kg0·60 per day (0·95 credible interval (CrI): 0·78–1·09) should be interpreted with caution. It was shown that boars’ ability to deposit protein was superior to that of barrows and gilts, as these had an estimated maximum PD (PDmax) of 250 g/day (0·95 CrI: 237–263), whereas the barrows and gilts had a PDmax of 210 g/day (0·95 CrI: 198–220). Furthermore, boars reached PDmax at 109 kg BW (0·95 CrI: 93·6–130), whereas barrows and gilts maximized PD at 81·7 kg BW (0·95 CrI: 75·6–89·5). At 25 kg BW, the boars partitioned on average 5–6% more of the ME above maintenance into PD than barrows and gilts, and this was progressively increased to 10–11% more than barrows and gilts at 150 kg BW. The Bayesian modelling framework can be used to further refine the analysis of data from metabolic studies in growing pigs.