Modelagem das exigências nutricionais de treonina em suínos das fases de creche, crescimento e terminação

Abstract

The objective of this work is to study the nutritional requirements of threonine in pigs in the nursery, growing and finishing phases, through systematic review and meta-analysis. Threonine is a nutritionally essential amino acid for pigs in the growth phases. This amino acid is important for animal growth and defense against pathogens acting on the immune system, in which it produces mucin and immunoglobulins. Therefore, the supply and estimation of threonine nutritional requirements for pigs is important, as its lack reduces animal performance. The database included 41 papers published between 1990 and 2020, totaling 4028 pigs, with 274 treatments. The selection criteria were: peer-reviewed papers, temporal space of 30 years, experiments with pigs in the nursery, growing and finishing phases, papers presenting at least 4 treatments (control + 3 levels of threonine), have performance data in the papers and composition of diets (percentage of ingredients). Data evaluation involved graphical and descriptive analyses, variance and covariance tests, mixed models, linear models, quadratic models, nonlinear models and machine learning. For the nursery and growing phases, the linear model was the best to estimate threonine requirements, where a daily gain of 38 and 18 grams was obtained for each gram of threonine ingested respectively. For finishing, the multiple model was the most appropriate, with a gain of 16 grams daily for each gram of threonine ingested. In the inverse form of estimation, for the nursery and growing phases, the linear model was the most appropriate, estimating 0.0103 and 0.0193 grams of threonine for each gram of daily weight gain respectively. At finishing, the multiple model was selected obtaining a Tre requirement of 0.01 grams for each gram of weight gain. In relation to the non-linear models, the reestimation obtained a gain of 45 grams for each gram of threonine ingested for the nursery phase and 26 grams for growing. For machine learning, the XGBOOST model presented the best model selection coefficients, in which the main predictor variables were weight gain, feed consumption and percentage of threonine in the diet. However, the use of machine learning was not suitable for estimating threonine nutritional requirements, due to limited data. It is concluded that through meta-analysis it was possible to estimate the nutritional requirements of threonine for pigs in the nursery, growing and finishing phases.