How to fit MET models with complex GxE variance structures for better outcomes

Mastering Multi-Environment Trials (MET) Analysis

The VSNi Team

17 April 2024

Multi-environment trial (MET) analyses are crucial for breeding programmes as they allow researchers to combine multiple trials or experiments into a single analysis. This helps to obtain breeding parameters, and understand genotype-by-environment dynamics and interactions, among others.

When fitting MET models, it is important to respect the individual personality of each trial site before putting them into a combined analysis. This includes carefully cleaning each dataset, performing spatial analysis (if possible), identifying and dealing with outliers, ensuring proper scaling, and proper referencing of control/check lines. Specialised statistically software such as ASReml-R can be used to fit these models, interpret variance components, extract genetic values and calculate heritability and type-B correlations.

For the single-stage analysis, researchers can build a mixed model with multiple fixed effects, random effects, covariates, and an appropriate complex error structure using all data simultaneously.

Alternatively, the two-stage approach involves first analysing each site individually, extracting predicted values (or adjusted means) together with its corresponding weights, and then using those results in a second step analysis focused solely on the modelling of the G x E interaction. Certain tools can streamline this workflow.

Throughout the process, it is important to consider best practices, such as the recommended minimum number of common genotypes between sites, proper handling of non-replicated trials, and strategies for addressing convergence issues with complex and/or unbalanced models.

Flexible G x E modelling approaches like the factor analytic structure can be particularly useful when dealing with a large number of trial sites. These structures allow researchers to balance model complexity with practical constraints. In addition, this structure allows for further understanding of the dynamics of this interaction.

By mastering MET analysis, you can gain valuable insights to support and improve your breeding programme. If you'd like to learn more, you can watch this webinar ‘How to fit MET models with complex GxE variance structures for better outcomes’ presented by Dr Salvador Gezan by clicking here.alt text