A virtual cell model comes out of the lab and points to the right drugs on real biopsies

Trained on millions of protein measurements, the model found on triple-negative breast cancer biopsies the treatments that proved effective in the patients.

A study published in Nature on September 9, 2026 describes a virtual cell model capable of evaluating the effectiveness of drugs directly on tissue samples. The system has been trained on several million protein measurements and tested on biopsies taken from patients with triple negative breast cancer.

In these experiments, the model identified the same molecules that were subsequently found to be effective in the patients involved. Threefold negative breast cancer accounts for 15 to 20% of breast cancer cases and remains difficult to treat due to the lack of conventional therapeutic targets; the perspective opened up by the authors is that of more personalized support.

Tiannan Guo, a proteomics specialist at the University of Westlake in Hangzhou and co-author of the study, points out what seems to him the decisive point: this is the first time that a virtual cell model leaves the laboratory to be tested in a clinical context.