A new machine-learning approach, developed by the University of Michigan, has created a "digital twin" capable of mapping real-time tumor metabolism in brain cancer patients. This technology aims to identify which treatment strategies, including dietary restrictions and specific drugs, are most likely to be effective against individual cases of glioma.
The model's accuracy was verified through comparisons with human patient data and experiments conducted on mice.
Technology Overview
The study, published in Cell Metabolism, builds on earlier research indicating that some gliomas can be slowed through dietary adjustments. Previously, determining which patients would benefit from restricting certain amino acids was challenging. The digital twin provides a method to make this distinction.
Additionally, the digital twin assesses the effectiveness of drugs designed to prevent tumors from producing DNA building blocks. It can identify if cancer cells can bypass a drug's effects by acquiring these molecules from their environment.
How the Digital Twin Operates
Developed to overcome challenges in mapping tumor metabolism inside the brain, the computer-based digital twin predicts individual brain tumor reactions to treatment. It integrates patient data obtained from blood draws, metabolic measurements of tumor tissue, and the tumor's genetic profile.
The system then calculates the metabolic flux, which represents the speed at which cancer cells consume and process nutrients. Researchers constructed a convolutional neural network, a type of deep learning model, and trained it using synthetic patient data, constrained by measurements from eight glioma patients.
Validation and Future Implications
Comparisons with different data from six of these patients demonstrated high accuracy in predicting metabolic activity. Mouse experiments further confirmed that dietary interventions slowed tumor growth only in mice identified by the digital twin as suitable candidates for the treatment.
The digital twin also accurately predicted tumor responses to the drug mycophenolate mofetil, identifying instances where tumors could utilize a "salvage pathway" to resist the drug's effects. These predictions were also confirmed through mouse experiments.
This technology intends to allow doctors to virtually test diets or drugs, facilitating personalized cancer care by focusing on treatments most likely to be effective and avoiding those a specific tumor is equipped to resist. The team has applied for patent protection and is seeking partners to commercialize the technology.
Support and Collaboration
The research received primary funding from the National Institutes of Health, specifically the National Cancer Institute, with additional support from various foundations. Contributors included researchers from the University of Alabama, Birmingham, and the Mayo Clinic.