Improving Protein Content Prediction in Wheat Grain

Abstract

The paper reports a study conducted with the goal of expanding the possibilities of using the previously obtained multiple nonlinear regression equation describing the dependence of protein content in wheat grain (Y, %) on the amount of crude gluten (X₁, %) and 1000-grain weight (X₂, g). It has been demonstrated that the original and three improved equations can be used in breeding practice to predict wheat grain protein content. The algorithm and the results of testing the predictive ability and prediction accuracy of the equations using independent data were presented. Analysis of the experimental data from 288 Russian and foreign publications, with a total number of observations n = 4604, obtained for more than 300 winter and spring soft and durum wheats in Russia and abroad (Albania, Belarus, Bulgaria, Egypt, Kazakhstan, Lithuania, Poland, Slovakia, and Ukraine) within the period from 1959 to 2020 showed that 755 observations, or 16.4%, fell beyond the variation limits accepted by GOST 10846-91: Method of Protein Determination. At the same time, the accuracy of protein content prediction reached 83.6%. The developed equations can be used for predicting protein content in almost all cases when the results of protein and crude gluten content analysis in wheat grain, as well as 1000-grain weight, are provided for actual or fixed moisture content or are recalculated for absolute dry matter.