Morphological and Pomological Characterization of Indigenous Pear Genotypes (Pyrus communis L. and P. syriaca Boiss.) and ANN-Based Prediction of Fruit Weight.

This study aimed to characterize the morphological and pomological diversity of 136 indigenous Pyrus communis L. and P. syriaca Boiss. genotypes collected from three major pear-growing provinces of Iran and to develop an artificial neural network (ANN) model for fruit weight prediction using morpho-pomological traits. A total of 40 quantitative and qualitative descriptors were evaluated. Phenotypic variation was analyzed using one-way ANOVA, Pearson correlation, principal component analysis (PCA), hierarchical cluster analysis (HCA), and multiple regression analysis. A multilayer perceptron (MLP)-based ANN model incorporating eight selected morpho-pomological variables was developed and validated using 10-fold cross-validation. Significant differences were observed among genotypes for all quantitative traits (p < 0.05), with 87.5% of the variables exhibiting coefficients of variation above 20%, indicating substantial phenotypic diversity. Fruit weight ranged from 7.46 to 139.69 g, whereas total soluble solids varied from 10.00% to 26.00%. Fruit weight showed strong positive correlations with fruit width (r = 0.96), fruit length (r = 0.86), mesocarp diameter, and flesh thickness, while total soluble solids were negatively associated with fruit size. PCA identified three major components representing fruit size, seed morphology, and leaf serration-ripening characteristics, enabling effective discrimination among genotypes. The ANN model achieved high predictive performance for fruit weight (R 2 = 0.9739, RMSE = 5.34 g, and MAE = 4.11 g). The indigenous pear germplasm exhibited considerable morphological and pomological variation with valuable breeding potential. The integration of multivariate analyses and ANN modeling provides an effective approach for genotype evaluation and accurate fruit weight prediction, supporting the selection of superior pear genotypes for breeding and cultivar development.