Decoding Glioblastoma Complexity Through Extracellular Vesicles, Organ-on-Chip Models, and Deep Learning.

Area of Science:

• Neuro-oncology
• Biotechnology
• Artificial Intelligence

Background:

• Glioblastoma (GBM) exhibits high heterogeneity and resistance, hindering effective treatment.
• Current experimental models lack the complexity to replicate GBM's dynamic features and generate multimodal data.

Purpose of the Study:

• To propose a novel organ-on-chip (OoC)-extracellular vesicle (EV)-deep learning (DL) framework for GBM research.
• To create a human-relevant experimental system that integrates GBM cells, microenvironment components, and EVs.
• To establish a roadmap for linking EV-mediated communication to GBM phenotypes and therapeutic responses.

Main Methods:

• Organ-on-chip systems incorporating patient-derived GBM cells, endothelial cells, astrocytes, pericytes, stromal cells, and immune components.
• Selective and temporal harvesting of extracellular vesicles (EVs) from defined microphysiological system compartments.
• Integration of imaging, barrier function, sensor, and EV-cargo data using modality-specific and multimodal deep learning (DL) architectures.

Main Results:

• The framework facilitates the generation of interpretable, multimodal datasets reflecting GBM's complexity.
• It enables the study of EV-mediated communication in relation to phenotypes like blood-brain barrier (BBB) disruption, invasion, immune reprogramming, and drug response.
• Technical challenges and considerations for BBB-on-chip systems, EV attribution, immune integration, and DL model selection are critically discussed.

Conclusions:

• The proposed OoC-EV-DL framework serves as an experimental roadmap, not an immediate clinical tool.
• This staged translational strategy aims to advance precision oncology in GBM by providing a biologically grounded approach.
• The framework addresses limitations in current GBM research by integrating dynamic biological features with advanced data analysis.