Disagreement Between Results Obtained from GWAS and Biparental QTL Mapping Suggests Epistatic Interaction

Two methods, genome-wide association study (GWAS) and biparental QTL mapping, can provide complementary information of QTL. In this study, we explored the genetics of yield-related traits in common buckwheat (Fagopyrum esculentum Monch.) through the two methods using populations derived from a single breeding population. Using a biparental population consisting of 178 plants, we developed a high-density linkage map comprised of 3,175 markers, and conducted biparental QTL mapping of ten yield-related traits: main stem length, number of nodes, stem diameter, timing of first flowering, number of flower clusters, number of primary branches, number of secondary branches, number of seeds per plant, 1000 seed weight, and test weight. Using a breeding population consisting of 165 plants, we conducted GWAS of the traits with the 1,378 markers. We compared the results obtained from the GWAS and the QTL mapping, and found that significant associations detected in both GWAS and QTL mapping were few in all of the traits, implying that loci controlling the traits were largely different between the two populations. Then, we searched significant epistatic interactions using the GWAS population. We detected a number of loci having epistatic interactions (p < 0.00001) with QTL detected in GWAS (p < 0.01) and QTL mapping (p < 0.001), and found that they showed the significance of overlap between GWAS and QTL mapping (p < 0.01) in five out of the ten traits. The results suggest that the disagreement between GWAS and QTL mapping is partly explained by the difference of population structures and epistasis.