EPCO-30. PROTEIN-BASED CLASSIFICATION REVEALS AN IMMUNE-HOT SUBTYPE IN IDH MUTANT ASTROCYTOMA WITH WORSE PROGNOSIS

Abstract Intertumoral heterogeneity poses significant challenges to the development of effective cancer treatment, particularly in the context of IDH-mutant astrocytoma. Although recent studies have revealed a couple of genomic features to stratify these tumors, treatment options are still limited, and clinical management of IDH-mutant astrocytoma remains controversial. To address these challenges, we leveraged an integrative spatiotemporal multi-omics strategy to profile IDH-mutant astrocytomas and discovered four reproducible protein-defined subtypes: AFM (adipogenesis/fatty acid metabolism), PPR (proliferative/progenitor), IME (immune/mesenchymal-enriched), and NEU (neuronal). PPR and IME subtypes were consistently associated with significantly worse prognosis, a trend confirmed across two independent cohorts, including one with two decades of follow-up. Molecularly, the PPR subtype is enriched for CDKN2A/B homozygous deletions and elevated mitotic activity, while the IME subtype is marked by presence of gemistocytic differentiation (GD), pronounced infiltration of cytotoxic T cells and plasma cells, and upregulation of interferon-stimulated genes. Single-cell and spatial transcriptomics revealed co-enrichment of mesenchymal-like glioma cells and lymphocytes within GD-rich niches. Among the upregulated genes in IME gliomas, GBP1 emerged as a candidate associated with malignancy, enhancing glioma cell proliferation in functional assays. Despite robust immune engagement, IME tumors exhibited poor survival, suggesting an immune dysfunction state possibly shaped by tumor–plasma cell interactions. To support clinical implementation, we developed GUIDE, an AI-enabled framework that integrates diverse omics modalities—including histopathology, transcriptome, proteome, and methylome—to identify the IME subtype with high accuracy, even in scenarios with partial data availability. Altogether, our study reveals a novel immune-hot, mesenchymal-enriched subtype of IDH-mutant astrocytoma with adverse prognosis and distinct spatial architecture. These findings provide new biological insights and offer a scalable AI-driven framework for precision classification, with implications for therapeutic targeting and disease monitoring in diffuse glioma.