A dataset for the application of the Bayesian AMMI (B-AMMI) model to identify high-yielding and stable barley...

One of the most crucial challenges for finding superior crop genotypes with wide or specific adaptation throughout multi-environment trials (MET) is analyzing genotype-by-environment interaction (GEI) and phenotypic stability. In this study, an existing MET dataset--including 18 advanced barley genotypes along with two check cultivars evaluated across 10 environments (five locations in Iran over two consecutive cropping seasons)--previously analyzed with classical Additive Main Effects and Multiplicative Interaction (AMMI) and GGE biplot models was comprehensively reanalyzed using a Bayesian framework. Unlike classical frequentist approaches that rely on fixed point estimates without quantifying estimation uncertainty, the Bayesian paradigm incorporates prior distributions and posterior probability density intervals, effectively mitigating overfitting and providing robust probabilistic credibility for GEI parameters. The evaluation integrated Bayesian AMMI (B-AMMI), Bayesian Weighted Average of Absolute Scores (BWAAS), the combined performance-stability index (BWAASY), and Mahalanobis stability distance. According to the B-AMMI2 biplot, the first two interaction principal component axes (IPCA1 = 41.87% and IPCA2 = 24.35%) jointly accounted for 66.22% of the GEI variance. This revealed significant environmental discrimination (E1, E3, E4, E5, E9, and E10) and delineated a mega-environment (E7, E8, and E9). While genotypes G7, G9, G12, G15, and G16 revealed specific adaptations, genotypes G8, G11, G17, G18, and G19 displayed high stability near the origin. Posterior estimations of BWAAS confirmed that G2, G10, G11, G18, and G19 were the most stable. Beyond confirming previously recognized elite lines (G3, G10, and G14), the Bayesian framework offered superior discriminative resolution: simultaneous selection via BWAASY and Mahalanobis stability distance effectively resolved ambiguity in candidate selection, revealing previously obscured top-performing stable genotypes (e.g., G3, G10, and G14) and confirming G3, G4, and G10 as possessing optimal compromises between high yield and low GEI sensitivity, while precisely classifying G14 as specifically adapted rather than broadly stable. Overall, re-evaluating MET data under the Bayesian AMMI framework provided a powerful and statistically sound probabilistic basis for breeding programs to increase selection accuracy.