Marker Prioritization With Different Fine Mapping Strategies Improves Genomic Selection Performance

Feature reduction of genetic markers and its impact on genomic selection has been studied extensively. Prior information borrowed from genome-wide association studies (GWAS) can improve genomic selection accuracy. However, the effect of integration of posterior inclusion probability (PIP) information from fine mapping algorithm and linkage disequilibrium (LD) on genomic selection remains less studied. Here, three marker prioritization strategies-genome-wide fine-mapping (GWFM), LD pruning (LDFM), and GWAS-informed LD pruning ((GWAS + LD)FM)-were evaluated, using Bayesian fine-mapping integrated with linear and nonlinear genomic selection frameworks, across 12 traits for three species. A consistent ranking of prediction performance (GWFM < LDFM < (GWAS + LD)FM) was observed across species and statistical models. Moreover, (GWAS + LD)FM achieved prediction accuracy (96% of maximal accuracy) with markedly minimal number of marker subsets. Gene ontology enrichment analysis showed that prioritized loci could converge on functional modules of biological importance. Independent validation using another fine mapping algorithm (FINEMAP) replicated the identical performance ordering. Genetic variants prioritized by LD and fine mapping reduced dimensional complexity and only were minimal number of marker subsets required for improved genomic selection accuracy. Fine mapping-informed dimension reduction provides novel insights and avenue for the enhancement of genomic selection performance.