Enhancing alfalfa breeding through genomic prediction with exotic germplasm resources

BACKGROUND: The adoption of molecular breeding methods and genomic selection in alfalfa breeding has lagged behind that of major crops such as corn and soybean. Discovery and introgression of native traits into new cultivars desired by growers requires breeders and pre-breeders to leverage genomic information from exotic germplasm sources into their breeding pipelines. RESULTS: We investigated the effects on genomic predictive ability when the composition and proportions of the germplasm used in the training and test sets vary in genomic prediction schemes. We used a set of 780 alfalfa samples collected from different geographic origins (CASIA, EURO, OTTM, and SIBR) plus U.S. breeding materials (Check) evaluated in a multi-year field trial for growth habit (GH), plant height (HGT), and plant vigor (VIG), and genotyped using the alfalfa 3 K DArTag panel. The genotyping results revealed distinct patterns of linkage disequilibrium across the five germplasm subgroups. Genomic prediction across the entire population outperformed predictions within individual subgroups. Further optimization was achieved by systematically varying the proportion of samples from each of the five subgroups in the training set (10-90%). Additionally, when increasing the number of test subgroup samples to the training set, the increased genetic relatedness between the training and test sets could be a major driver of improved predictive performance. CONCLUSIONS: Overall, these results indicate that successful genomic prediction in alfalfa requires not only expanding training set sizes but also strategically incorporating highly diverse germplasm and evaluating newly introduced base populations to maximize predictive ability.