Multivariate Dissection of Trait Domains and Trade-Offs in Provitamin-A Cassava Genotypes

Improving provitamin‑A cassava requires a clear understanding of how key agronomic and nutritional traits respond to seasonal variation and how these traits interact within a multivariate structure. The objectives were to evaluate agronomic and nutritional traits across seasons, apply multivariate analyses to uncover trait domains, and investigate inter‑trait relationships to guide breeding strategies. Forty‑two provitamin‑A cassava accessions were evaluated using a split‑plot randomized complete block design, with genotype as the main‑plot factor and harvest time as the subplot factor, replicated across seasons. Eight traits were measured, with emphasis on DM, FYLD, and TC. Multivariate analyses provided deeper insight into trait structure. Principal component analysis and factor analysis identified distinct trait structures: PCA revealed four trait domains, while factor analysis uncovered three latent trait groupings–Root Productivity (ML2), Vegetative and Compositional Diversity (ML3), and Harvest Efficiency (ML1). Traits such as RTWT and HI exhibited near‑zero uniqueness, reinforcing their roles as anchor indicators of productivity and efficiency, while TC displayed exceptionally high uniqueness, confirming its independence from yield domains and its regulation by distinct metabolic pathways. Biplots highlighted genotype dispersion and trait loadings, hierarchical clustering grouped accessions by combined agronomic and nutritional performance, and a chord diagram confirmed a tightly linked yield complex alongside a separate nutritional domain. Variance component estimates revealed contrasting levels of genetic and environmental influence. DM showed moderate genotypic variance and no significant seasonal effect, underscoring its stability and reliability for processing quality. FYLD exhibited low genotypic variance, high residual variance, and a highly significant season effect, confirming strong environmental sensitivity. TC displayed high genotypic variance and a significant seasonal effect, suggesting that carotenoid accumulation is both genetically controlled and environmentally modulated. The integration of mixed‑model variance partitioning with multivariate and network‑based analyses revealed clear differences in trait stability, genetic control, and inter‑trait relationships. These findings identify DM as a robust trait for selection, FYLD as highly environment‑dependent, and TC as a promising target for biofortification with manageable environmental sensitivity. The results provide a comprehensive model for index‑based selection strategies, guiding breeders toward “stacked” ideotypes that combine high yield potential, stable dry matter content, enhanced carotenoid accumulation, and efficient resource allocation to meet industrial, nutritional, and food security needs across cassava‑growing regions.