Genotype x environment interactions and implications for associations among soybean traits

An option into guide selection strategies in soybean breeding programs is to investigate associations between the main traits studied by breeders. A question that is faced is the possible influence of genotype x environment (GE) interactions on correlations among traits. Path analysis allows, in addition to measuring correlation, a more detailed study of cause / effect relationships among traits. We measured the effect of the GE interaction in the association between agronomic traits in soybeans. The experiments were carried out during two crop seasons, in three municipalities of Minas Gerais, Brazil, using a randomized complete blocks design, with 35 treatments. We evaluated number of grains, number of pods (NP), number of grains per pod, grain yield, weight of 100 grains, lodging score, plant height, height of first pod insertion, and full maturity. The data were submitted to variance analysis and path analysis. We also calculated the sum of ranks index in three ways, considering different economic weights. The correlation between grain yield and other traits, as well as direct and indirect effects, varied in magnitude and direction, considering individual and multi-environmental analyses, demonstrating the influence of the GE interaction. For example, considering the individual analysis for Lavras in 2015/16, the correlation between grain yield and NP was -0.0534 and was not significant at the 5% level by the Mantel test, and the direct effect of NP on grain yield was 0.1717, while in the joint analysis of all environments, the correlation between these traits was 0.3674 and was significant, and the direct effect of NP on grain yield was -0.0520. The direct effect as an economic weight in the sum of rank index can be more efficient in selecting cultivars when compared to a simple correlation; this could become a useful strategy for the selection of multiple traits.