Abstract While somatic drivers, including mutations, fusions, and copy number alterations, are well-characterized, transcriptome-wide splicing remains elusive. Its proteomic impact in pediatric tumors is particularly poorly understood, highlighting the critical need for comprehensive characterization from patient samples. Here, we quantified splicing across 1,415 primary CNS tumors from the Children’s Brain Tumor Network (CBTN) and identified both histology-specific and shared splicing patterns. Among the four major splicing types (single exon, alternative 5’ splice site, alternative 3’ splice site, and retained intron), single exon events dominated, with medulloblastomas, low-grade, and high-grade gliomas exhibiting the highest number of uniquely recurrent variants. Consensus clustering based on percent-spliced-in (PSI) values revealed 12 distinct groups across histologies (n = 17). Clusters were not exclusively histology-specific; for instance, one cluster contained all major medulloblastoma subtypes, while high-grade gliomas were broadly distributed. Splicing burden was also significantly correlated with spliceosome pathway activity, highlighting mis-spliced splicing factors as drivers for these changes. Gene set variation analysis demonstrated cluster-specific programs reflecting distinct oncogenic processes and pathways. Moreover, 64% (n = 7,960) of differential splicing variants impacted coding regions, including phosphorylation and acetylation binding sites, suggesting that tumors generate patient-specific proteomes shaped by unique splicing patterns. To explore this further, we plan to integrate RNA and proteomic data (n = 967) modalities to link tumor-specific splicing events to proteomic profiles. This dual-level characterization will reveal unique molecular features in individual patients, offering insights into precision therapeutics tailored to splicing-driven proteomes. Additionally, it may uncover tumor-specific splicing events and mechanistic alterations that drive cancer progression, advancing both personalized medicine and broader translational discoveries within pediatric brain cancers. Finally, maximizing the impact of these findings, we will build scalable infrastructure and robust data integration frameworks into existing Kids First architecture, addressing key challenges in harmonizing and disseminating these rich datasets to the research community.
Naqvi et al. (Fri,) studied this question.