Linear and nonlinear models for measuring protein nutritional quality.

Abstract

Several widely used linear methods for assessing protein nutritional quality were compared to a nonlinear model. Male, weanling Sprague-Dawley rats were fed diets containing 2--31% casein, peanut and wheat gluten protein for 4 weeks. One control group received a protein-free diet and another was killed at the start of the experiment. Nitrogen intake and animal response in terms of body weight and nitrogen were measured for each rat. PER, NPR, NPU, linear regression (LR) slopes and saturated kinetics (SK) parameters were calculated from the data. The linear models provide a single estimate of protein quality, while the SK model can estimate quality at any point along the intake-response curve. At maintenance (delta Body Weight or delta Body Nitrogen = 0) the SK model ranked peanut and gluten protein higher relative to casein than the linear methods, while at higher protein intakes the SK model gave lower relative rankings to those proteins. The SK model closely fit the experimental data over the entire intake-response range but did not converge to a meaningful solution when an insufficient quality of quantity of protein was included in the diet. Unlike the linear models, the SK model failed to distinguish statistically between the quality of proteins in several cases.