Designing superior crosses in cassava using genomic mating to boost yield and genetic diversity

Identifying superior cassava ( Manihot esculenta Crantz) crosses through phenotypic evaluations is costly and inefficient due to cassava's long breeding cycles and low flowering rates. In this study, we applied genomic mate selection and optimum contribution selection using additive, dominance, and directional dominance models to enhance prediction accuracy and optimize cross design in cassava. A total of 3391 clones were evaluated across 49 multi‐environment trials for fresh root and shoot yield, dry matter, starch content, and plant architecture. Directional dominance effects improved predictive ability for all traits except starch, emphasizing the role of nonadditive effects in cassava breeding. Genomic mating enhanced predicted gains for yield and quality traits, while reducing predicted values for plant architecture, aligning with selection for compact ideotypes. Although optimum contribution selection effectively controlled inbreeding, it reduced the number of selected crosses, reflecting a trade‐off between diversity and gain. Parent‐wise cross‐validation confirmed that directional dominance models consistently produced higher predicted means for fresh root and shoot yield and dry matter content. The most promising crosses were identified based on a multi‐trait selection index and usefulness criteria, integrating mean performance and within‐family variance. Our results demonstrate that combining directional dominance modeling with genomic mating tools increases breeding efficiency, identifies crosses with superior predicted performance, and supports ideotype‐based breeding. This strategy offers a cost‐effective approach for accelerating genetic gain while maintaining diversity in cassava improvement programs.