Evaluation of text mining-guided SNP weighting methods for genomic prediction of fatty acid traits in Hanwoo cattle

Fatty acid composition in longissimus dorsi muscle is a key determinant of beef quality in Hanwoo cattle, yet accurate genomic prediction of them remains challenging due to their complex polygenic architecture. This study evaluated whether integrating text mining-derived biological priors into weighted genomic best linear unbiased prediction framework could improve accuracy and heritability for seven fatty acid traits: myristic, palmitic, stearic, oleic, linoleic, eicosenoic, and arachidonic acids. 1,161 Hanwoo cattle genotyped with the Illumina 777 K SNP chip were used. Candidate genes were identified through text mining, and corresponding SNPs formed a biologically informed marker subset. Weighted genomic relationship matrices were constructed using GWAS-derived and BayesR-derived SNP effects and were compared with original full-marker and LD-pruned GBLUP. Original full-marker and LD-pruned GBLUP yielded accuracies of 0.021–0.463 and 0–0.453, respectively, and heritabilities of 0.0003–0.424 and 0–0.432. GWAS TM-weighted GBLUP yielded accuracies of 0.204–0.514 and heritabilities of 0.035–0.747, whereas BayesR TM-weighted GBLUP yielded accuracies of 0.347–1.007 and heritabilities of 0.109–0.969 across traits, although the accuracy exceeding one for one trait indicated potential instability under highly heterogeneous BayesR-derived weights. These findings suggest that integrating text-mined biological priors with SNP weighting can enhance genomic prediction for fatty acid traits in Hanwoo cattle when appropriately implemented.