Selection of superior soybean genotypes based on multi-trait using MGIDI and FAI-BLUP indices in multi-environmental trials

Soybean (Glycine max [L.] Merril) holds global significance as a legume, both economically and as a food source. Its global production for the 2023 harvest has been estimated at about 401 million tons. The main destinations for soybean grains are human consumption and livestock industry due to its high protein and oil content. Genetic breeding goals to select superior genotypes so that each year the production of this crop surpasses productivity records while consuming fewer resources. For this scenario, a profile of an ideotype based on traits desired is constructed to select genotypes that come closest to ideal. Therefore, the use of robust models and indices that can weigh various traits for this selection is crucial. We selected purelines based on multi-trait and stability through the REML/BLUP approach. For this, two indices were calculated: factor analysis and ideotype-design (FAI-BLUP Index) and the Multi-trait Genotype-Ideotype Distance Index (MGIDI). The dataset consists of 106 purelines evaluated across 15 environments (where environment represents the combination of location and cropping season) in the state of Mato Grosso do Sul. The experimental design employed a randomized complete block design with three replications. The experimental area consisted of five rows measuring 12 m each, spaced at 0.5 cm intervals, totaling na area of 30 m2. To determine those that achieved the desired performance for the crop, multi-traits were evaluated, including seed yield (kg ha-1) (SY), number of flowers per plant (NFP), height (H), and lodging rate (LR). The ideal treatment was defined as the one that would provide the desired values for all traits. The ideal genotype should provide: high SY, a higher NFP, tall stature, and a lower LR. Thus, for LR, lower values were defined, and for the other traits, higher values were desired. Assuming a selection intensity of 15%, the sixteen purelines selected using MGIDI were: L006, L051, L017, L069, L002, L005, L080, L033, L019, L070, L079, L100, L083, L011, L037, and L020. And for FAI, they were: L006, L069, L051, L017, L019, L002, L021, L020, L033, L097, L037, L011, L005, L100, L080, and L039. Thirteen purelines coincided in both indices when the contribution graph was generated. Finally, a graph using the view of the strengths and weaknesses of the MGIDI index was created to identify the ideal genotypes for each trait based on a multi-trait framework. The four traits were grouped into factors (FA) as follows: In FA1, three characteristics were grouped (NFP and H with positive loads and LR with negative loads); And in FA2, GR with positive loads. Thus, L019 was considered ideal for FA1, and L051 for FA2, being the most productive among all selected purelines.