Evaluation of biochemical profiles in walnut (Juglans regia L.) genotypes and predictive modeling using advanced machine learning algorithms

Abstract

Biochemical parameters such as phenolic compounds, tocopherols and polyunsaturated fatty acids (omega-6 and omega-3 etc.) found in walnuts support cardiovascular health with their high antioxidant content, reduce inflammation and improve the blood lipid profile and show positive effects such as lowering LDL cholesterol, increasing HDL and reducing oxidative stress. This study investigates the application of machine learning algorithms to predict the biochemical properties of 10 walnut (Juglans regia L.) genotypes selected from the Kahramanmaraş region and three commonly cultivated walnut cultivars (Maraş-18, Chandler, and Franquette). Plant material was collected between 2020–2021, and analyses were conducted on biochemical components, including polyphenols, antioxidant capacity, protein, oil, fatty acids, sugars, and organic acids using HPLC, GC, and spectrophotometric methods. The highest total phenolic content was found in genotype 46KM-5 (402.6 mg GAE/100 g), and the highest antioxidant capacity in genotype 46KM-13 (73.77%). The 46KM-9 genotype exhibited the highest oil content (71.60%) and the highest total sugar content (3.27%), while sucrose was the dominant sugar. Regarding fatty acids, genotype 46KM-19 had the highest linoleic acid content (60.88%), and genotype 46KM-21 showed the highest oleic acid content (20.88%). Malic acid emerged as the dominant organic acid, with the highest content observed in genotype 46KM-9 (6.55%). In addition, various machine learning algorithms such as such as decision tree, support vector regression, K-nearest neighbor (KNN), gradient boosting, multi-layer perceptron, and linear regression (LR) have been used to assess the biochemical properties of genotypes using hyperparameter optimization. KNN algorithm generally provided high performance in prediction processes by exhibiting very low error rates. This study's results align with previous research, highlighting walnut genotype variation and nutritional benefits. Machine learning predicts key traits, aiding breeders and consumers in health-focused production and better genotype selection for improved cultivation.