Integrating Genetic Mapping and Genomic Prediction to Elucidate the Genetic Architecture of Fusarium Ear Rot Resistance in Tropical Maize

Fusarium ear rot (FER) caused by Fusarium verticillioides is a major constraint on global maize production. The genetic basis of FER resistance is not yet fully understood, and the development of effective breeding strategies for improving FER resistance is still a critical priority. In the present study, a collection of 254 CIMMYT tropical maize lines genotyped with 955,690 high-quality SNPs was used to conduct genome-wide association studies (GWAS), complemented by QTL (quantitative trait locus) mapping in two recombinant inbred line populations. Additionally, genomic prediction (GP) exploring various statistical models and SNP selection schemes was implemented to optimize predictive accuracy for improving FER resistance. The broad-sense heritability estimates of FER resistance were 0.69–0.86 in the CML panel across six environments and 0.39–0.69 in the two RIL populations. At a p-value threshold of 2.61 × 10−7, GWAS identified 18 SNPs significantly associated with FER resistance across six environments, and in single environment analyses, their phenotypic variance explained (PVE) values ranged from 0.68 to 13.75%, with 13 SNPs exceeding a PVE of 5%. At a p-value threshold of 1 × 10−5, an additional 37 SNPs were detected, clustering within seven environmentally stable regions identified in at least two environments. Furthermore, 13 haplotype blocks exhibiting significant phenotypic differences were identified within these stable regions, with PVE values ranging from 2.39 to 15.24%, 9 of which exceeded 5%. QTL mapping in the two RIL populations revealed 27 moderate-effect QTLs at a LOD threshold of 2.5, including four detected repeatedly across environments, though only one QTL overlapped with the GWAS-identified region. Moderate genomic prediction accuracies of FER severity were achieved across models, with GBLUP and BayesB outperforming other models, and the prediction accuracies of these two models in the three populations were all around 0.5. Integrating the significant SNP set from genetic mapping results with a 100-SNP background set enhanced the stability of cross-population predictions. These results implied that FER resistance in tropical maize is controlled by multiple genomic regions with small-to-moderate genetic effects, whereas the consistency of genomic regions detected by GWAS and QTL mapping is low. Genomic prediction incorporating regions identified across different genetic backgrounds emerges as a promising tool for accelerating FER resistance breeding.