The genetic architecture of seed composition in soybean is refined by genome-wide association scans across multiple populations

Soybean oil and meal are major contributors to world-wide food production. Consequently, the genetic basis for soybean seed composition has been intensely studied using family-based mapping. Population-based mapping approaches, in the form of genome-wide association (GWA) scans, have been able to resolve loci controlling moderately complex quantitative traits (QTL) in numerous crop species. Yet, it is still unclear how soybean's unique population history will affect GWA scans. Using one of the populations in this study, we simulated phenotypes resulting from a range of genetic architectures. We found that with a heritability of 0.5, ∼100% and ∼33% of the 4 and 20 simulated QTL can be recovered, respectively, with a false-positive rate of less than ∼6×10(-5) per marker tested. Additionally, we demonstrated that combining information from multi-locus mixed models and compressed linear-mixed models improves QTL identification and interpretation. We applied these insights to exploring seed composition in soybean, refining the linkage group I (chromosome 20) protein QTL and identifying additional oil QTL that may allow some decoupling of highly correlated oil and protein phenotypes. Because the value of protein meal is closely related to its essential amino acid profile, we attempted to identify QTL underlying methionine, threonine, cysteine, and lysine content. Multiple QTL were found that have not been observed in family-based mapping studies, and each trait exhibited associations across multiple populations. Chromosomes 1 and 8 contain strong candidate alleles for essential amino acid increases. Overall, we present these and additional data that will be useful in determining breeding strategies for the continued improvement of soybean's nutrient portfolio.