LEVERAGING A FUNCTIONAL AND EVOLUTIONARY SIGNIFICANCE SCORE FOR VARIANT PRIORITIZATION IN GENOMIC PREDICTIONS OF VNN RESISTANCE IN EUROPEAN SEA BASS

Disease outbreaks still represent a serious threat to aquaculture. Conventional selective breeding for complex traits such as disease resistance, relying on estimated breeding values, can be expensive and difficult to implement. In contrast, genomic selection using high-density SNP genotyping provides a more efficient strategy to enhance genetic resistance. Although whole-genome sequencing (WGS) provides opportunities for genomic selection, the inclusion of non-informative variants can introduce noise and increase computational burden. To address this limitation, we used the Functional-And-Evolutionary Trait Heritability (FAETH) score (Xiang et al., 2019) to prioritize sequence variants for genomic prediction of viral nervous necrosis resistance (VNN) in European sea bass. A total of 990 juvenile sea bass were phenotyped for body length (mm) and binary VNN mortality after a 29-day VNN experimental challenge test. Parents and 40 experimental fish were whole-genome sequenced (6,072,853 SNPs), enabling WGS imputation for all experimental individuals, genotyped with the Med-Fish SNP array (27,740 SNPs). The FAETH score was calculated for WGS variants by averaging per-variant heritability estimates for two traits (body length and VNN mortality) across several functional and evolutionary partitions: ATACseq/ChIP-seq data, Ensembl annotations, and genomic islands of differentiation between Atlantic and Eastern Mediterranean European sea bass lineages. Variant subsets (1%, 5%, 10%, 25% and 50% SNPs with the highest FAETH scores) were tested in prediction of genomic estimated breeding values for VNN mortality using Bayesian threshold models (GIBBSf90+/BLUPf90+). Eight scenarios, differing in LD-pruning filters (no filtering, 0.99, 0.95 and 0.80 thresholds) and in the inclusion or exclusion of the known major QTL for VNN resistance on chromosome 3 (Mukiibi et al., 2025) were considered. Model accuracies were assessed in a 2-fold cross-validation, minimizing genetic relatedness between training and testing sets. In all scenarios, high-ranking variants showed from 2.03% to 8.90% increase in prediction accuracy compared to the bottom 25% low-ranking variants. This study represents the first application of a functional and evolutionary significance score to prioritize SNPs in an aquacu