Prediction of antioxidant capacity in faba bean from individual phenolic constituents

Abstract

AbstractThe relationship between individual phytochemical constituents and overall antioxidant capacity or total phenolic content (TPC) is poorly understood in faba bean. This study used a range of linear and nonlinear regression techniques to investigate whether the antioxidant capacity and TPC of 60 faba bean samples (flour and methanolic extracts) could be predicted from 12 individual compounds (10 common polyphenols and 2 alkaloid glycosides) measured in the same samples. Nonlinear regression using machine learning with a Radial Basis Function showed the best performance for antioxidant and TPC prediction across all sample types, while multiple linear regression allowed moderately accurate predictions in most sample matrices. Improved performance metrics were seen for the methanolic extracts compared to the flour samples. The strongest predictors of antioxidant activity in the multiple linear regression models were protocatechuic acid, p-hydroxybenzoic acid, and ferulic acid, suggesting that these compounds are particularly important contributors to the high antioxidant activity of faba bean. Understanding the relationship between individual constituents and the antioxidant capacity may help food technologists and plant breeders develop faba bean products with maximal health benefits.