Researchers have developed a cutting-edge AI platform that can design tailor-made protein components to train a patient’s immune cells to target and destroy cancer. Published in Science, the study marks a major milestone in precision oncology. For the first time, scientists have shown that they can design proteins on a computer to redirect immune cells toward cancer cells using peptide-MHC (pMHC) molecules—an approach that drastically reduces the time to discover effective treatments from years to just 4–6 weeks.
A New Vision for the Immune System
“We’re essentially giving the immune system a new set of eyes,” said Associate Professor Timothy P. Jenkins of the Technical University of Denmark (DTU) and senior author of the study. Current therapies rely on finding specific T-cell receptors from patients or donors—a slow and labor-intensive process. In contrast, the new AI-driven platform rapidly creates molecular keys, or “minibinders,” that can direct immune cells to recognize and attack cancer cells.
Targeted Missiles Against Tumors
As reported by medicalxpress, the platform, jointly developed by DTU and the Scripps Research Institute in the U.S., addresses a critical challenge in cancer immunotherapy: how to attack tumor cells without harming healthy tissue. Normally, IMPAC-T cells recognize cancer through specific protein fragments (peptides) displayed on pMHC molecules. However, due to the wide variation in T-cell receptors among individuals, designing personalized therapies has proven difficult—until now.
AI-Designed Minibinders Show Potent Results
To demonstrate the platform’s potential, the researchers focused on NY-ESO-1, a cancer antigen present in multiple tumor types. The AI successfully designed a minibinder that tightly attached to NY-ESO-1 pMHC molecules. When inserted into IMPAC-T Cells, this new construct—named IMPAC-T cells—efficiently targeted and killed cancer cells in lab tests.
Postdoctoral researcher Kristoffer Haurum Johansen at DTU shared his excitement: “It was incredibly exciting to take these minibinders, created entirely on a computer, and see them work so effectively in the lab.”
Custom Solutions for Unique Cancer Targets
The team also tested their AI system on a new target discovered in a patient with metastatic melanoma. Once again, the platform generated effective minibinders, proving that the method can be adapted for novel and individualized cancer targets.
Built-In Virtual Safety Screening
To enhance safety, the researchers incorporated an AI-based screening process to predict whether the designed proteins might mistakenly bind to healthy tissue. This virtual filter helped eliminate high-risk molecules before any lab testing, significantly reducing the chance of side effects.
“Precision is everything in cancer therapy,” said DTU Professor and co-author Sine Reker Hadrup. “By screening for cross-reactions early in the design phase, we increased both the safety and success rate of our treatments.”
Path to Clinical Trials
While the results are promising, researchers estimate that it may take up to five years before this AI-designed therapy reaches clinical trials. Once approved, the process would resemble CAR-T therapy, currently used for leukemia and lymphoma.
How the Treatment Will Work
Patients will have blood drawn at the hospital. Immune cells will then be extracted and engineered in a lab with the AI-designed minibinders. These modified T cells will be reintroduced into the patient’s body, where they act like precision-guided missiles—locating and destroying cancer cells with high accuracy.
This AI-powered method has the potential to revolutionize cancer treatment by making personalized, safe, and effective therapies available on a much larger and faster scale.