A new AI-based virtual cell model was developed to predict effective breast cancer drugs for triple-negative breast cancer (TNBC) by analyzing proteomics data from tumor cells.

To train the AI, researchers tracked 5,585 protein groups at baseline and at 6, 24, and 48 hours after drug exposure to capture dynamic cellular changes.

The model was trained on more than 38 million protein measurements from 18 breast cancer cell lines, including 16 TNBC lines, and tested against 63 FDA-approved antitumor drugs and 59 drug combinations.

Analysis of 3,651 proteins in tumor biopsies from 501 TNBC patients before chemotherapy showed the model accurately predicted the clinical outcomes of the treatments those patients received.

The AI achieved an 88% accuracy in predicting cellular responses to 81 drugs not included in the training data, demonstrating strong predictive capability for unseen drugs in TNBC.

The study points toward more personalized TNBC treatment by using the virtual cell model to guide drug selection, representing a first clinical use of such a model.