Resolving variants of uncertain significance in neurofibromatosis: An integrated approach combining deep learning and minigene assays.

Neurofibromatosis (NF) comprises genetic disorders mainly caused by pathogenic variants, yet its phenotypic and genotypic heterogeneity complicates diagnosis. We analyzed clinical and genomic data from 97 NF patients using targeted panels, whole-exome sequencing (WES), and whole-genome sequencing (WGS) from June 2020 to October 2024. Variants were classified according to established guidelines, and their distribution across protein domains was evaluated using Bayesian multinomial logistic regression. Deep-learning prediction tools and minigene splicing assays were applied to assess variants of uncertain significance (VUS). Sixty-nine variants were identified in NF1, NF2, and LZTR1, including 22 novel ones. In NF1, pathogenic deletions were enriched in non-domain regions, while substitutions predominated in domain regions, though without phenotype-specific associations. Two of three VUS were predicted and experimentally confirmed as pathogenic. One case achieved molecular diagnosis only through WGS after negative WES results. This study expands the mutational landscape of NF genes, underscores the diagnostic advantage of WGS, and demonstrates the effectiveness of advanced predictive and functional tools for VUS interpretation.