Genomic selection for maize quality traits

Genomic selection (GS) has emerged as an efficient breeding strategy for improving complex quantitative traits in maize (Zea mays L.). By utilizing genome-wide molecular marker information, GS enables the prediction of genomic estimated breeding values (GEBVs) and facilitates early selection of superior breeding materials. Maize quality traits, including protein, oil, starch, lysine, tryptophan, provitamin A carotenoids, micronutrients and kernel characteristics, exhibit diverse genetic architectures. While some nutritional quality traits are governed predominantly by major genes or majoreffect loci and have been successfully improved through marker-assisted selection (MAS), including markerassisted backcrossing and gene pyramiding, many other quality traits are polygenic and controlled by numerous loci with small effects. For such complex traits, GS offers an advantage by simultaneously capturing genomewide genetic variation and estimating the cumulative effects of multiple loci. Prediction accuracy is influenced by training population size, genetic relatedness, marker density, trait heritability, phenotyping quality, population structure and genotype × environment (G×E) interaction. Various statistical approaches, including genomic best linear unbiased prediction (GBLUP), ridge regression BLUP (RR-BLUP), Bayesian models and machine-learning methods, have been evaluated for genomic prediction in maize. Recent studies have demonstrated the potential of GS for improving protein, oil, starch, kernel texture and other nutritional and processing-quality traits. Integration of GS with MAS, high-throughput phenotyping, genomewide association studies, and multi-trait and multienvironment prediction can further enhance selection efficiency. Thus, GS should be viewed as a complementary approach to MAS, with particular value for accelerating the improvement of complex, polygenic maize quality traits and developing nutritionally superior and end-use-specific maize varieties.