# Multi-model-based validation of multi-environment trial results in horse gram (Macrotyloma uniflorum Lam. Verdc.)

## TL;DR

This study validates the performance of promising horse gram genotypes across multiple environments using mixed models and identifies stable, high-yielding varieties.

## Contribution

The study integrates fixed and random effect models to validate genotype performance in horse gram, revealing breeding potential and method-specific superiority.

## Key findings

- Genotype G1 showed the highest yield relative to environmental maximum (YREM) of 1.00, indicating stability across environments.
- Genotypes G1, G3, G22, G27, and G8 were identified as superior performers based on multi-trait selection index and BLUP analysis.
- Selection differential analysis confirmed the superiority of selected genotypes for key traits like yield and plant characteristics.

## Abstract

Analysis of genotype-by-environment interactions (GEI) is critical for evaluating the yield and stability of genotypes in multi-environment experiments (METs). Either fixed models (AMMI and GGE biplots) or random effect models (Linear mixed models: LMM) are utilised to estimate GEI. From our preliminary METs using fixed models, a few potential horse gram mutant genotypes (G1, G3, G25, & G27) were identified. To validate this finding, confirmatory METs were performed during the 2023 (E1, E2, & E3) and 2024 (E4, E5, & E6) cropping seasons using LMM and the yield relative to the environmental maximum (YREM) index at pulses experimental farm, Coimbatore (E1 & E4), sugarcane research station, Melalathur, Vellore (E2 & E5), and farmer’s field, Krishnagiri (E3 & E6). The traits positively and significantly correlated with yield were utilized for LMM and YREM analyses. The YREM value ranged from 0.55 to 1.00. The genotype G1 exhibited the highest YREM of 1.00, indicating that the crossover GEI effect did not affect its yield potential. After the likelihood ratio test (LRT), BLUP analysis was performed and identified the promising genotypes G1, G3, G22, G25, G27, and G8 based on higher predicted mean values. The genetic variance components (GVC) estimation showed low R² GEI (0.056 to 0.539), high ‘As’ (0.993 to 0.785), low to high rge (0.127 to 0.758), and medium to high H2 (0.068 to 0.863). The WAASB biplot identified potential genotypes (G1, G8, G27, G22, and G3) based on yield potential and stability. The multi-trait selection index (MTSI) recognized the genotypes G1 (1.192), G22 (1.250), G3 (1.319), and G8 (1.841) as superior performers. The combined YREM and LMM results revealed the significant breeding potential of G1, G3, G22, G27 and G8 for exploitation, while G22, G25, and G8 exhibited method-specific superiority. The selection differential analysis affirms the superiority of these genotypes for the tested traits (NC-25.70%, NPC-14.20%, NP-28.80%, NS-9.94%, YD-30.40%, & DM-0.05%). In conclusion, the validation METs reveal a slight variation among the findings between fixed and random effect models, emphasising the importance of integrating multiple analytical approaches for an effective GEI analysis in horse gram.

The online version contains supplementary material available at 10.1038/s41598-025-25093-2.