AI Tool ProteinTalks Predicts Personalized Breast Cancer Drug Responses

Not all cancer cells respond to drugs alike, making the right treatment hard to match to the right tumor. A new AI tool called ProteinTalks tackles this by tracking how a cancer cell's proteins change over time after a drug is applied, to predict whether the drug will work. Described in a Nature study led by Rui Sun, ProteinTalks functions as an AI "virtual cell" — a computer simulation of a living cell's behavior.

The team captured more than 38 million protein measurements across 5,585 proteins from 18 breast cancer cell lines treated with 63 anticancer drugs approved by the FDA, the US drug regulator, and 59 two-drug combinations, measuring protein levels and survival before treatment and again at 6, 24 and 48 hours. That time-resolved data let ProteinTalks learn how protein levels shift as a drug takes effect, unlike most AI models, which treat cell behavior as a single snapshot.

ProteinTalks outperformed AI models based on gene activity and traditional machine-learning methods. It accurately predicted responses to 81 anticancer compounds it had never seen during training, and revealed 4 drug pairs that worked better together than alone against triple-negative breast cancer — an aggressive, hard-to-treat cancer lacking hormone receptors. Though trained only on breast cancer, it also predicted lung, colorectal, pancreatic and melanoma cancer responses.

Tracking these protein changes pinpointed AKR1C3 as a key driver of drug resistance: knocking down the protein restored cancer cells' sensitivity to the chemotherapy drug docetaxel, a finding confirmed with a cytotoxicity assay, a lab test that measures how well a treatment kills cells. It also analyzed protein patterns in 501 triple-negative breast cancer tumors to group patients by recurrence risk and survival outcome.

Screening more than 3,000 repurposed drugs against patient-derived organoids — miniature tumors grown from patient tissue — ProteinTalks identified personalized drug candidates that could kill tumor cells at far lower doses than standard chemotherapy, and found new synergistic drug combinations.