Pathogenicity Reassessment and Novel Variant Discovery in Inherited Retinal Disease through Population-Scale Genomics in the United Arab Emirates

BackgroundPathogenicity classifications for rare disease variants are largely derived from clinically ascertained cohorts, in which penetrance is often assumed to be near-complete. As genomic sequencing expands into population-scale screening, particularly in underrepresented populations, both the penetrance of known variants and the relevance of novel or population-enriched variants require systematic evaluation. Inherited retinal diseases (IRDs), genetically heterogeneous and clinically variable, represent an ideal setting in which to assess how ascertainment-based pathogenicity assignments translate to real-world population risk. MethodsWe analyzed whole-genome sequencing data from 504,000 participants in the Emirati Genome Program, including 426,382 individuals with linked longitudinal electronic health records and 75,184 genetically reconstructed family units. IRD phenotypes were defined through a phenome-wide association approach to derive enriched retinal ICD-10 code sets. Variants in 339 IRD-associated genes were assessed using a population-based framework integrating enrichment testing, penetrance tiering, and family-based concordance across pathogenic or likely pathogenic variants (P/LP), variants of uncertain significance (VUS), and novel variants not previously classified in clinical variant databases, yielding a population-calibrated IRD panel. ResultsAmong 24,100 variants observed in individuals with IRD-compatible genotypes (259 P/LP, 11,888 VUS, and 11,953 novel variants), 533 variants were prioritized. P/LP variants showed the highest proportional enrichment (15.4% vs. 3.4% of VUS and 3.2% of novel variants), with the strongest population-level signals being P/LP variants in KCNV2 and ABCA4, both markedly enriched in the Emirati cohort relative to global reference data and consistent with their predominance among affected individuals at regional ophthalmology clinics. Restricting prioritization to P/LP variants alone would have captured only 37 variants; inclusion of VUS and novel variants expanded the panel more than 14-fold. Conversely, 9.3% of P/LP variants lacked population-level enrichment, including 3.1% with null penetrance despite adequate carrier representation. Family-based concordance supported 404 of 533 prioritized variants, defining the population-calibrated panel; 91.8% of these were VUS or novel. ConclusionsPenetrance of IRD-associated variants is heterogeneous and frequently lower than inferred from clinical cohorts. Population-scale calibration identified clinically relevant VUS and novel variants, including population-specific risk alleles absent from global databases, while revealing limited real-world support for a subset of established P/LP classifications. These findings underscore that assertion-based pathogenicity classifications require recalibration against real-world disease expression before deployment in population-scale genomic screening, particularly in underrepresented populations.