REML implementations of kernel-based multi-trait, multi-environment genomic prediction models скачать в хорошем качестве

REML implementations of kernel-based multi-trait, multi-environment genomic prediction models

As breeding programmes increasingly rely on genomic prediction across multiple environments and traits, modelling genotype-by-environment interactions accurately becomes both more challenging and more critical. In this webinar, presented by Killian Melsen (Biometris, Wageningen University & Research), we explore how the integration of environmental data into linear mixed models can significantly improve predictive accuracy. Breeding datasets that span multiple years or locations are often sparse, and many genotypes are not tested in every environment. A small number of check varieties are typically used to provide connectivity, but this limited overlap can make it difficult to estimate genetic correlations between environments. The result? Poor model convergence, unreliable G×E estimates, and reduced prediction accuracy. In this webinar, we demonstrate how to use ASReml-R to improve model accuracy in these settings by incorporating environmental covariables, drawn from weather stations or open-source satellite data, into your linear mixed models. These covariables can be used in both linear and non-linear forms, with either environment-specific or averaged genetic variances, to better capture genotype-by-environment interactions. We also show how to extend this approach to handle multiple traits or management practices using unstructured modelling, making it easier to integrate phenomic, enviromic, and genomic data within the linear mixed model. All models are implemented using ASReml-R, with practical examples and flexible, ready-to-use code for any number of environments, traits, or management conditions with no manual tuning required. Presented by: Killian Melsen, PhD student, Biometris, Wageningen University & Research Killian is part of the mathematical and statistical methods department at Wageningen University & Research. After gaining hands-on experience in plant breeding through several internships during his BSc in applied biology, he pursued an MSc in plant sciences at WUR, specialising in plant breeding and quantitative methods. Now a PhD researcher at Biometris, Killian focuses on the integration of diverse omics datasets into plant breeding using linear mixed models