Local genomic estimates provide a powerful framework for haplotype discovery

Quantitative trait loci (QTL) discovery studies on diversity panels or breeding populations typically use genome-wide association studies (GWAS) to estimate marker effects. For plant and animal breeding applications, researchers increasingly recognize the potential benefits of identifying superior haplotypes (markers in linkage disequilibrium; LD) rather than relying on single markers, as traditional approaches inefficiently account for cumulative signals from incomplete LD with QTL or split effects when multiple markers are in high LD with QTL. Using the genomic prediction framework, the local GEBV (localGEBV) method was developed in animal breeding and has been adopted in crop haplotype mapping studies; however, no study has thoroughly quantified the utility of this method or systematically compared outcomes to traditional GWAS approaches. Here, we characterized a strategy to group markers in chromosomal segments based on LD (haplotype blocks or haploblocks), computed localGEBV as a linear contrast of marker effects within each haploblock, and utilised the variance of localGEBV to enhance QTL discovery compared to traditional GWAS. Marker effects for localGEBV were estimated with ridge-regression best linear unbiased prediction (rrBLUP) and BayesR, with results compared to two common GWAS approaches. Using the barley row-type trait, we demonstrated that localGEBV improved QTL discovery and phenotypic prediction compared to single markers. Furthermore, localGEBV results were robust to the choice of prior marker assumptions and blocking parameters, enabling flexibility in fine or broad-scale QTL mapping. Overall, our findings establish localGEBV as a haplotype-based strategy capable of leveraging localized genomic effects to improve QTL discovery and, potentially, genomic selection.