Abstract
BACKGROUND: Prenatal diagnosis of X-linked hydrocephalus caused by variants in the L1CAM gene is often complicated by the identification of Variants of Uncertain Significance (VUSs). This study showcases an accelerated diagnostic workflow using artificial intelligence (AI) to rapidly interpret a novel missense variant for a family with a history of the disorder. METHODS: We performed exome sequencing (ES) on a male fetus with significant sonographic brain anomalies from a 29-year-old pregnant woman. To efficiently analyze the resulting VUSs, we used the AI tool AlphaMissense to predict their pathogenicity and prioritize them for validation. The top candidate variant was then assessed via Sanger sequencing for co-segregation across eight maternal relatives. The structural impact of the mutation was visualized using the AlphaFold 3 model. RESULTS: Exome sequencing identified four VUSs. AlphaMissense predicted only one, L1CAM c.1228C>G (p.His410Asp), as 'likely pathogenic'. Subsequent Sanger sequencing confirmed that this variant co-segregated perfectly with the disease phenotype in the family. Based on this strong genetic evidence, the variant was reclassified from a VUS to 'Likely Pathogenic'. Structural modeling revealed that the p.His410Asp substitution disrupts a critical salt bridge, likely compromising protein stability. CONCLUSION: Our two-step approach-using AI for rapid VUS prioritization followed by targeted Sanger validation-proved to be a highly efficient strategy. It provided a definitive and clinically actionable diagnosis that facilitated genetic counseling and enabled the family to pursue Preimplantation Genetic Testing (PGT). This workflow significantly enhances the power of genomic testing in the prenatal setting.