Integrating multi-trait genomic selection with simulation strategies to improve grain yield and parental line selection in rice

Inclusion of correlated secondary traits in the prediction of primary trait in multi-trait genomic selection (GS) models can improve the predictive ability. Our objectives in the present investigations were to (i) evaluate the effectiveness of multi-trait and single-trait GS models for the higher predictive ability and (ii) compare the breeding potential of parental lines selected based on phenotype and GS for grain yield in rice. We used phenotype data of five correlated traits as secondary traits evaluated to predict the grain yield, a primary trait. Yield related functional markers were used for prediction. Breeding populations were simulated using the best parents selected through GS and phenotype based selection. Results suggest that the multi-trait model resulted in higher predictive abilities (0.82 for grain yield) than single-trait models (0.76 for grain yield) and parents selected through GS have potential to produce superior progenies. We conclude that the use of a multi-trait GS approach is advantageous over single-trait models, and the GS also help selecting potential parents for developing improved populations. The results of the study have potential scope for improving quantitative traits using GS in rice.