# Multi-trait and multi-environment genomic prediction enhances yield components improvement in durum wheat

## TL;DR

This study shows that combining multiple traits and environments in genomic prediction improves durum wheat yield stability in changing climates.

## Contribution

The novelty lies in demonstrating the effectiveness of multi-trait–multi-environment genomic prediction models for durum wheat improvement.

## Key findings

- MTME models achieved highest prediction accuracies under specific cross-validation and sowing scenarios.
- Incorporating functional allele data improved predictions for morpho-phenological traits like heading date and plant height.
- MTME models effectively captured genotype performance across Mediterranean environments, aiding climate-resilient breeding.

## Abstract

Durum wheat [Triticum turgidum L. ssp. durum (Desf.) Husn.] is a staple crop for the pasta and semolina industries, particularly in Mediterranean and semi-arid regions where climate variability poses major challenges to yield stability. This study evaluates the performance of single-environment (SE), multi-trait (MT), multi-environment (ME), and multi-trait–multi-environment (MTME) genomic prediction models across seven key traits, such as grain number per spike, grain weight per spike, number of spikelets per spike, spike length, spike weight, heading date, and plant height. Using genomic (G) and target gene-based (G2) relationship matrices with two cross-validation scenarios (CV1 and CV2), MTME models achieved the highest prediction accuracies, particularly under CV2 and sowing-by-season grouping. Modeling G2 information improved predictions for morpho-phenological traits (i.e. heading date and plant height), confirming the utility of functional allele data for capturing gene effects. MTME models effectively leveraged inter-trait and inter-environment covariance, providing biologically realistic predictions of genotype performance across simulated Mediterranean environments. These findings establish MTME genomic prediction as a powerful and scalable framework for climate-resilient durum wheat improvement, supporting predictive and data-driven breeding pipelines aimed at enhancing genetic gain and stability across years and environments.