Bayesian probabilistic selection index in the selection of common bean families

Abstract

Selecting progenies evaluated in different seasons, locations, and years is a challenge for breeders in plant breeding programs. This is because different environmental conditions can lead to differential expression of genes involved in controlling traits of interest, resulting in the genotype × environment (G × E) interaction. Utilizing the G × E interaction appropriately can enhance the selection of multiple traits when progenies are evaluated in different environments. Probabilistic Bayesian models have shown the ability to consider the effects of the G × E interaction to calculate the risk of recommending a given candidate genotype for selection. Therefore, the objectives of this study were to propose a selection index based on probabilistic Bayesian models and to apply it to a selection of bean families. To this end, 380 common bean families from the third recurrent selection cycle of the carioca common bean breeding program at the Federal University of Viçosa (UFV) were evaluated over four environments for the following traits: grain yield (GY), commercial grain appearance (CGA), and plant architecture (PA). Based on the proposed Bayesian probabilistic selection index (BPSI), which ranks the multitrait superior performance of families across environments, and a selection intensity of 10%, 12 superior families were identified. These families had a higher probability of superior performance in all environments for the traits GY (28%), CGA (52%), and PA (62%) simultaneously compared to the check cultivars. Compared to other indices, the BPSI selects families with lower sum ranks, for example, top positions have a probability of superior performance across environments, and with at least 42% of families that do not match other indices. The BPSI selection index showed promise for selecting families in a common bean breeding program.