Cancer therapy is undergoing a dramatic transformation with the rise of artificial intelligence. Scientists are now using AI to rapidly develop customized immune therapies, designing synthetic proteins that train the body’s defenses to detect and destroy tumors. Unlike conventional methods—which often take years to identify natural immune matches—this new approach can deliver results in just weeks.
A study published in Science reports that researchers from the Technical University of Denmark and the Scripps Research Institute have built an AI platform capable of targeting tumor-related proteins such as NY-ESO-1, a molecule found in many cancers and long viewed as a promising immune trigger. Using the system, the team designed a “minibinder” protein that locks onto NY-ESO-1, boosting the immune system’s ability to recognize and eliminate cancer cells.
The process involved engineering immune cells with these AI-generated proteins, producing what the researchers call IMPAC-T cells. In laboratory experiments, these modified cells showed strong precision in identifying and killing tumor cells expressing NY-ESO-1. According to study co-author Kristoffer Haurum Johansen, the minibinders performed strikingly well under controlled lab conditions.
What makes the platform even more powerful is its flexibility. Beyond known cancer markers, the AI tool successfully created minibinders for an unexplored protein from a patient with metastatic melanoma. This demonstrates its potential to tackle cancers that previously lacked effective immune targets, opening the door to therapies designed around the unique markers of each individual patient. As DTU associate professor Timothy P. Jenkins explains, the platform effectively gives the immune system new molecular “keys” to unlock and destroy tumors at unprecedented speed.
One of the biggest challenges in designing immune therapies is avoiding harm to healthy tissue, since many cancer markers resemble normal proteins. To address this, the researchers built in a “virtual safety screen.” Each minibinder was tested in silico against a wide range of healthy cell proteins to rule out dangerous cross-reactions before moving into lab experiments. This step significantly reduces the risk of harmful side effects and enhances the likelihood of safe clinical applications.
Despite the promising lab results, human trials are still several years away. Jenkins estimates that clinical testing could begin in about five years, following a process similar to existing treatments for blood cancers, where immune cells are harvested, engineered, and reinfused into patients. The difference is that AI accelerates and personalizes the design phase, potentially overcoming the barriers that have limited immune therapies for solid tumors.
The ability to generate entirely new, cancer-targeting proteins from a digital model marks a milestone in immunotherapy. Within weeks, scientists can now move from a computer-generated blueprint to functional immune cells that rival traditionally developed ones in both safety and effectiveness.
As this technology advances, it promises to make personalized treatments more accessible and precise, expanding options for cancers once considered untreatable and for patients with rare or unique mutations. AI-driven design is not just enhancing biological discovery—it is redefining how future cancer therapies will be created.