Long-read sequencing reveals widespread novel splicing and neojunction-derived neoantigens in nasopharyngeal carcinoma

The widespread transcriptomic diversity driven by alternative splicing contributes to all hallmarks of cancer and represents a critical source of neoantigens for personalized immunotherapy. However, unlike other major malignancies, the full repertoire of alternative splicing events in nasopharyngeal carcinoma (NPC) remains underexplored. Here, we employ long-read sequencing (LR-seq) to generate a high-resolution, isoform-level transcriptomic atlas from a cohort of 14 NPC tumor samples and four immortalized nasopharyngeal epithelial cell lines. We identify a substantial number of full-length novel transcripts (22,687; ~44.38%), which reveal diverse splicing patterns and previously unannotated splicing events. By integrating short-read RNA-seq data to quantify isoform expression, we discover a subset of novel transcripts that are differentially expressed between tumor samples and immortalized nasopharyngeal epithelial cell lines. Furthermore, LR-seq enables precise identification of chimeric read-through fusion transcripts, such as CLDN15-FIS1 and FOXRED2-TXN2 . Finally, we develop a computational framework - Tumor-specific splicing neoantigen detection (TS-SNAD) to predict neoantigens originating from novel exon-exon junctions (neojunctions) in tumor-specific novel transcripts. Using this framework, we identify neojunction-derived neoantigens and experimentally validate the immunogenicity of selected neoantigens restricted by HLA-B*40:01. These neojunction-derived peptides constitute a new class of noncanonical neoantigens with significant potential for developing personalized cancer vaccines for NPC.