AI 'virtual cell' model helps predict which drugs may work for a hard-to-treat cancer

On September 12, 2026, Medical Xpress reported on a study in the journal Nature about ProteinTalks, an AI system described as a virtual cell. It aims to predict how tumor cells from triple-negative breast cancer will respond to different drugs. Nature notes that this type makes up 15 to 20% of breast cancer cases and is harder to treat because hormone and targeted therapies cannot act on its cells. The team, led by Rui Sun, trained the model on more than 38 million measurements of 5,585 proteins, molecules that carry out many jobs inside cells. They tracked breast cancer cells before and at several points after treatment with 63 approved cancer drugs and 59 two-drug combinations. The model pointed to four drug pairs that worked better together, and in tests with tumor samples from patients it showed promising results in picking drugs that proved effective when those patients received them. The reports describe research findings and do not say the model is being used to choose treatments in hospitals yet. Why it matters: The authors say this approach raises the possibility of more personalized cancer care, and it is a clear example of AI being used as a tool for science. Source: Medical Xpress — https://medicalxpress.com/news/2026-09-ai-virtual-cell-protein-dynamics.html Source: Nature — https://www.nature.com/articles/d41586-026-02845-2