Sponsored by
Researchers at the University of Michigan developed a machine-learning based “digital twin” program that maps how individual brain tumors use nutrients. Published in Cell Metabolism, the approach integrates patient blood data, tumor genetics, and limited metabolic measurements taken during surgery, including isotope tracing, to predict tumor-specific metabolic dependencies. Some tumors rely on particular amino acids or pathways and may be slowed by dietary changes or drugs, while others can bypass these restrictions.
The researchers validated the model using data from glioma patients who received labeled glucose during surgery and confirmed predictions in mouse models. The digital twin accurately identified which tumors would respond to the drug mycophenolate mofetil, while also distinguishing tumors that could evade the drug by using alternative “salvage” pathways to obtain DNA-building materials.
This is exciting work, and we hope to see further development and validation!