Bayesian estimation of genotype-by-environment interaction in sorghum variety trials

摘要

Genotype × Environment Interaction (GEI) plays an important role in identifying genotypes for high and stable yield for broad and specific adaptations. It continues to be a challenging issue among plant breeders and agronomists when conducting crop performance trials across diverse and unpredictable environments. Normally, the analysis of GEI is carried out under the frequentist paradigm, even though ongoing crop improvement programs gather information on genotypic and experimental error parameters that could be incorporated using a Bayesian approach. The objective of this paper was to estimate, for sorghum ( Sorghum bicolor ) in Sudanese environments, genotypic and GEI variances, heritability, genetic advance attributable to selection, and genotype means using Bayesian and frequentist approaches. Eighteen genotypes of sorghum were evaluated in randomized complete block designs with four replicates in six environments, during 2009/10 - 2011/12, at South-Gedarif and North-Gedarif in Sudan. Priors were obtained from a previous set of multi-environment trials in sorghum during 2006/7 – 2008/9 at Rahab, Sudan. Estimates of heritability and genetic advance under the Bayesian approach were higher than those under the frequentist approach. Precision of means of genotypes and heritability estimates were also higher under the Bayesian approach. The Bayesian approach provides a wider coverage for statistical inference and incorporates prior information with the likelihood of current data. For this approach, an illustrative step-by-step procedure is presented and recommended for use in statistical analysis of crop genotypes from multi-environment trials.