Abstract Tumor heterogeneity is increasingly recognized as a major obstacle to therapeutic innovation across neuro-oncology. Gliomas typically exhibit a spectrum of possible genetic mutations, reflecting multiple, potentially complex interactions across distinct molecular pathways. Predicting disease evolution and prescribing individually optimal treatment plausibly requires statistical models complex enough to capture the intricate genetic structure underpinning oncogenesis. Here we formalize this task as the discovery of distinct patterns of connectivity within network representations of the interactions between genetic loci. Evaluating multi-institutional clinical, genetic, and outcome data from 7847 consecutive patients over a 14-year period, we employ Bayesian stochastic block modelling to reveal a hierarchical network structure of tumor genetic interactions spanning (molecularly confirmed) IDH-wildtype glioblastoma, IDH-mutant oligodendroglioma and astrocytoma. We show that this structure not only identifies the patients’ diagnosis, but predicts individual survival with greater fidelity (Cox’s proportional hazard cross-validated (CPH-CV) concordance = 0.78; Bayesian 12, 24 and 36-month survival models = R2 0.31, 0.40 and 0.46, respectively) than the latest WHO tumor classification (CPH-CV concordance = 0.77; Bayesian 12, 24 and 36-month survival models = R2 0.26, 0.36 and 0.39, respectively) or linear models of genetic features (CPH-CV concordance = 0.76; Bayesian 12, 24 and 36-month survival models = R2 0.26, 0.35 and 0.37, respectively). For example, a diagnosis of IDH-wildtype glioblastoma yielded median survivals varying from 289 days in patients with the signature of marked EGFR amplification, MGMT methylation, and histone mutations, to 425 days in those with the signature of TERT mutations, histone and EGFR wild-types, and variable MGMT methylation. These findings suggest network-based analysis can reveal distinct signatures of survival with better prognostic fidelity than current gold standard diagnostic categories.
Read full abstract