Selection indexes based on linear‐bilinear models applied to soybean breeding

Recently developed selection indexes provide solutions for plant breeding, using linear-bilinear models that consider factors as fixed or random. This work aimed to compare the multitrait selection indexes based on factor analysis and ideotype-design (FAI-BLUP), GGE biplot, and grain yield x trait index (GYT), and proposes the use of predicted genetic values together with the GYT index (best linear unbiased prediction used in grain yield(*)trait index, GYT-BLUP). In addition, this work indicates the best index to select superior soybean [Glycine max (L.) Merr.] genotypes, closer to the ideotype. Data from 35 homozygous soybean lines and four checks, were obtained from trials conducted in six locations in the southern region of Brazil in the 2014/2015 crop season. The grain yield, yield components, morphological and grain composition were evaluated. Phenotypic data were used for GGE biplot and GYT analysis, using the software GGE biplot. Genetic values were predicted with mixed models considering genotype and location as random and fixed effects, respectively. Thus, genetic values were used in GYT-BLUP and FAI-BLUP indexes. These methods were compared by Spearman's rank correlation. Genetic gains obtained by indexes and traits were estimated. Soybean lines L1 and L22, and cultivars C3 and C4 were selected based on their performance for multiple traits, for indexes used. Thus, we suggest to combined FAI-BLUP and GYT-BLUP indexes. The GYT-BLUP has a high importance for grain yield, which was related to all other traits. FAI-BLUP gave similar weights for all traits. So, combining different approaches can provide better answers to breeders.