Adult and paediatric malignant glioma (GBM and pHGG) remain among the most difficult-to-treat cancers with 5-year survival rates of 300, including pre-post treatment samples) with matched controls (n>300) and exceptionally long-term surviving GBM patients (n~140), in which various tumour-host niches will be studied in how they respond to I/O perturbations and lead to improved clinical outcome. This will be empowered by deploying an UNCAN-compatible data lake, to which incremental data collection will be used to further refine the machine learning models, while proposing novel treatment options. This action is part of the Cancer Mission cluster of projects on “Understanding (tumour-host interactions)".