Genetic divergence of common bean lines for agronomic traits by hierarchical methods considering multicollinearity

Genetic divergence for agronomic traits in common bean lines can be analyzed with and without multicollinearity and by different hierarchical methods, which can lead to errors in the interpretation of results obtained from dendrograms. The objectives of this study were to evaluate the variability and genetic divergence of common bean lines for agronomic traits and to analyze the clustering pattern of hierarchical methods obtained with and without multicollinearity. The study involved data from three experiments in which 12 Mesoamerican bean genotypes were evaluated for 17 agronomic traits. Genetic divergence was analyzed through four hierarchical methods, considering two degrees of multicollinearity: severe (with multicollinearity) and weak (without multicollinearity). Ten of the 17 agronomic traits showed a significant genotype × environment interaction effect, indicating response variability of the genotypes in different growing environments. When performed with and without multicollinearity, the hierarchical methods resulted in varied clustering patterns for the genotypes regarding agronomic traits. When analyzed with and without multicollinearity, the furthest neighbor and Ward methods revealed low cophenetic correlation coefficients (≤ 0.69), explaining why the generated dendrograms do not adequately represent the variation observed between common bean genotypes. The unweighted pair group method with arithmetic mean (UPGMA) and nearest neighbor methods implemented without multicollinearity provide a high cophenetic correlation coefficient (0.76). However, the nearest neighbor method is more discriminative in allocating the common bean genotypes into different groups, allowing for a better recognition of differences regarding agronomic traits.