Statistical methods for genetic evaluation and selection of parents and hybrids of grain sorghum

This is the authors' abstract. We don't add key points for this paper.

• Bayesian multi-trait multi-environment model were used for sorghum breeding. • Bayesian genetic correlation among traits differed across tested sites. • This model might be successfully used in the genetic selection of sorghum genotypes. Grain sorghum is one of the most important cereals and is grown in arid and semi-arid regions due its high photosynthetic efficiency and tolerance to water deficit. Genotype-by-environment interactions and multi-trait selection are the main challenges in grain sorghum breeding. The objective of this study was to compare statistical methods for genetic evaluation and selection of parents and hybrids of grain sorghum. To this end, we conducted two trials and measured flowering, plant height and grain yield traits. Bayesian and frequentist multi-trait multi-environment (MTME) models were fitted through Markov Chain Monte Carlo (MCMC) and Restricted Maximum Likelihood/Best Linear Unbiased Prediction (REML/BLUP), respectively. The Additive Index was used to perform the multi-trait selection. The Bayesian inference is more flexible, and our results suggest that the Bayesian MTME model outperform the frequentist MTME model and should be preferred for genetic evaluation. The Additive Index can be used for genetic selection of parents and hybrids of grain sorghum.