Abstract
The variability in genetic variance and covariance due to genotype × environment interaction (G×E) can hinder genotype selection accuracy, especially for complex traits. This study analyzed G×E interactions in cassava to identify stable, high-performing genotypes and predict agronomic performance in untested environments using factor analytic multiplicative mixed models (FAMM) within multi-environment trials (METs). We evaluated 22 cassava genotypes for fresh root yield (FRY), dry root yield (DRY), shoot yield (ShY), and dry matter content (DMC) across 55 Brazilian environments. FAMM was applied to estimate genetic values and environmental loads, revealing significant genetic variance, especially for FRY (0.16-0.92) and broad-sense heritability ([Formula: see text]) above 0.70 in advanced yield trials. In joint analyses, analytic factor FA4 explained over 88% of genetic variation for all traits despite high G×E and data imbalance. Positive genetic correlations were found between environments for ShY and DRY (0.99 and 1.0, respectively), while FRY and DMC showed negative correlations (-0.82 and -0.95). Latent regression analysis identified hybrids adaptable to a range of environments, as well as genotypes suited to specific conditions. Moderate correlations between environmental covariables (rainfall, altitude, solar radiation) and FA model loadings suggest these factors contribute to high G×E interactions, notably for FRY. The FAMM model provided a robust approach to G×E analysis in cassava, yielding practical insights for breeding programs.