A practical framework for predicting splicing single nucleotide variants in exome sequencing

Splicing variants significantly contribute to Mendelian disorders, yet predicting their pathogenicity remains challenging. To address this issue, we developed a framework that simplifies the evaluation of pathogenic splicing single nucleotide variants (SNVs) while following ACMG/AMP guidelines and ClinGen recommendations established in 2023. Our system simplifies the 2023 ClinGen criteria by assigning a priority score (ranging from -10 to 14) to SNVs in open reading frame regions. Validation using pathogenic splicing SNVs from the Human Gene Mutation Database and common SNVs from gnomAD demonstrated superior discrimination compared to SpliceAI alone (area under the receiver operating characteristic 0.991 versus 0.983, P = 2.11 × 10⁻²³). When applied to 1257 patients with unresolved diagnoses after exome sequencing, our framework identified pathogenic splicing variants in COL2A1, PDHA1, MECP2, and JAKMIP1 and suggested potential candidate disease-causing genes, UBN1 and NFE2L1. This method enhances the detection of splicing variants in exome sequencing.