Development of a FER0.4K SNP array for genomic predication of Fusarium ear rot resistance in maize

Fusarium ear rot (FER) caused by Fusarium species severely reduces grain yield and quality of maize. Genome prediction (GP), a promising tool for quantitative trait breeding in plants and animals, uses molecular markers for capturing quantitative trait loci and predicting the genetic value of candidates for selection. In the present study, different subsets of markers and statistical methods for GP accuracy were tested in diverse inbred populations for FER resistance using a five-fold cross-validation approach. The prediction accuracy increased with an increase in the number of random markers; however, an increase in number beyond 10K did not increase the prediction accuracy. The prediction accuracy of selected markers was higher than that of random markers, and 500–1000 selected markers had the highest prediction accuracy, beyond which it slowly decreased. Although there was no difference among statistical methods when using selected markers at high prediction accuracy, significant differences were observed when using random markers. On this basis, a liquid chip named FER0.4K (liquid chip for genomic prediction of FER) containing 381 SNPs was developed for low-cost, high-throughput genotyping, with a prediction of approximately 0.82. The statistical method of genome prediction was compiled into a web-based, easy-to-use statistical analysis software using the “shiny” package in R. In summary, this study provides a foundation for FER resistance breeding in maize and offers new insights into the genetic improvement of other complex quantitative traits in plants.