A spatiotemporal state-inference framework for adaptive immunotherapy in glioblastoma

Area of Science:

• Neuro-oncology
• Immunotherapy
• Computational Biology

Background:

• Immunotherapy has improved outcomes in many solid tumors but shows limited efficacy in glioblastoma (GBM).
• This is partly due to the immunosuppressive tumor microenvironment and a mismatch between fixed treatment schedules and the evolving tumor-immune ecosystem.
• Understanding GBM's temporal and spatial immune evolution is crucial for effective treatment.

Purpose of the Study:

• To introduce the GBM Immune-Spatiotemporal Feedback Loop (GBM-ISFL) framework for adaptive immunotherapy in GBM.
• To conceptualize immunotherapy as a longitudinal process of sensing, state inference, intervention, and feedback.
• To propose novel computational tools for noninvasive inference of tumor-immune states.

Main Methods:

• Review of single-cell and spatial multi-omics, radiologic assessments, and liquid biopsy studies.
• Development of a four-phase atlas of GBM evolution and a patient-specific Critical Transition Window.
• Application of spatiotemporal graph neural networks (STGNNs) for inferring latent tumor-immune states from multimodal data.

Main Results:

• A conceptual framework (GBM-ISFL) for adaptive immunotherapy based on biologic phase rather than fixed chronology.
• Identification of a patient-specific Critical Transition Window within GBM evolution.
• Proposal of STGNNs for noninvasive inference of tumor-immune states from serial multimodal data.

Conclusions:

• The GBM-ISFL framework offers a testable approach for adaptive, state-informed precision immunotherapy in GBM.
• This framework reframes GBM immunotherapy around biologic phase, moving beyond fixed treatment schedules.
• Further research is needed to validate this approach and integrate it into clinical practice, as it is not a current standard of care.