Brain cancer in silico modelling - the roadmap ahead

Brain cancer comprises a diverse group of aggressive malignancies, often associated with a poor prognosis and a pressing need for more effective therapeutic strategies. The complexity of brain tumours arises from their highly dynamic and heterogeneous nature, involving intricate biological processes and interactions across multiple scales. Addressing these complexities requires an integrative approach that incorporates various levels of biological organisation to develop more accurate and predictive models of brain cancer. Cutting-edge techniques, including advanced imaging and omics technologies, provide invaluable spatial, molecular, and morphological data over time. These innovations offer deeper insights into tumour progression, microenvironment interactions, and treatment responses, enhancing the foundation for computational modelling. Mathematical and computational models are crucial tools for synthesizing these multiscale data, enabling cancer evolution simulation, treatment efficacy exploration, and therapeutic strategies optimisation. In this review, we organised a bibliographic dataset (2000-2025) along three major axes-mathematical methodologies, data-generating technologies, and biological hallmarks. Leveraging a novel machine learning-based approach to navigate this structured landscape, we uncovered emergent patterns and key trends in the evolution of computational modelling strategies for brain tumours. Our analysis reveals a temporal shift toward data-driven mathematical methods and high-throughput data technologies, particularly AI-based approaches and imaging- or omics-based inputs. At the same time, comparatively less emphasis is placed on incorporating biological complexity. Data-driven and mechanistic modelling approaches, often seen as distinct, are converging toward synergistic frameworks that combine biological interpretability with adaptive prediction. To realize this potential, sustained efforts are needed to deepen our biological understanding of brain tumours and to generate robust, diverse datasets that can drive both mechanistic insight and predictive power.