New AI-driven cancer treatment builds immune defenses in weeks

In a major step toward faster and more personalized cancer treatment, scientists have unveiled a powerful AI-driven method that creates custom-designed immune therapies in just weeks. Instead of relying on the slow and complex task of finding natural immune system matches, this new approach uses artificial intelligence to create precise proteins that can guide immune cells directly to cancer targets.

The method works by designing small, synthetic proteins—called minibinders—that stick to specific molecules on cancer cells. These molecules, known as peptide-major histocompatibility complexes (pMHCs), present protein fragments from within the cancer cell on its surface.

Normally, the immune system’s T cells use their T-cell receptors to recognize these fragments. But finding and testing the right T-cell receptors from patients or donors can take months or even years. Thanks to advanced generative models and simulations, scientists can now create these minibinders entirely in a computer, test them virtually, and produce a working prototype in the lab within just 4 to 6 weeks.

In a study published in the journal Science, researchers from the Technical University of Denmark (DTU) and the Scripps Research Institute described how they used the AI platform to target a well-known cancer protein called NY-ESO-1. This protein is found in many types of tumors and is known to activate the immune system.

The researchers trained the AI to design a minibinder that could attach tightly to the NY-ESO-1 pMHC structure. Once the protein was created, it was inserted into immune cells in the lab. These modified cells, called IMPAC-T cells by the team, showed a strong ability to recognize and kill cancer cells carrying the NY-ESO-1 marker.

“It was incredibly exciting to take these minibinders, which were created entirely on a computer, and see them work so effectively in the laboratory,” said Kristoffer Haurum Johansen, a postdoctoral researcher at DTU and co-author of the study.

The AI system doesn’t just work on known targets. The scientists tested it again using a different pMHC target, this time from a metastatic melanoma patient. This target, called RVTDESILSY/HLA-A*01:01, had not been structurally mapped before. Even without a known structure, the platform successfully generated minibinders that matched the new cancer protein, opening the door to designing treatments for previously untargetable cancers.

This is a huge advantage in precision medicine. Instead of relying on limited available data or difficult-to-access immune cells, scientists can now use digital models to create effective therapies for targets unique to each patient’s cancer. “We are essentially creating a new set of eyes for the immune system,” said Timothy P. Jenkins, associate professor at DTU and the study’s senior author. “Our platform designs molecular keys to target cancer cells using the AI platform, and it does so at incredible speed.”

One of the most challenging parts of developing new immune therapies is making sure they don’t attack healthy cells. Some cancer markers closely resemble proteins found in normal tissues. If the therapy binds to these by mistake, it can cause severe side effects.

To prevent this, the research team added a “virtual safety check” to their process. Using computer simulations, they tested each minibinder against a wide range of pMHCs found on healthy cells. This allowed them to rule out potentially harmful designs before any lab testing began.

“Precision in cancer treatment is crucial,” said Sine Reker Hadrup, professor at DTU and co-author of the study. “By predicting and ruling out cross-reactions already in the design phase, we were able to reduce the risk associated with the designed proteins and increase the likelihood of designing a safe and effective therapy.” This predictive step is key to ensuring that these therapies can one day be used safely in people. By eliminating risky binders early, researchers can focus resources only on the most promising and safest molecules.

While the lab results are promising, scientists still need to move carefully. Jenkins estimates it will take about five years before the first clinical trials in humans can begin. Once trials are underway, the treatment will likely resemble existing methods used for certain blood cancers.

In those therapies, doctors collect a patient’s blood, extract immune cells, and modify them in the lab. The modified cells are then returned to the patient’s body, where they hunt down and kill cancer cells.

With the new AI method, the process becomes faster and more personalized. Instead of waiting to find the right immune cell or matching receptor, doctors can use digital tools to design a perfect fit. This could be a game-changer for patients with solid tumors, where current immune therapies have had limited success.

By tailoring the therapy to each patient’s tumor markers, doctors may soon be able to treat cancers that were previously untreatable. In addition, the AI method makes it easier to explore rare or individual mutations that traditional therapies overlook.

The ability to design effective cancer-targeting proteins from scratch has long been a goal in immunotherapy. This breakthrough shows that it’s now possible to go from a digital blueprint to a working immune therapy in just a few weeks.

The team’s work also proves that these synthetic minibinders can function just like natural receptors in guiding T cells to their targets. When tested in the lab, the engineered immune cells were just as deadly to cancer as those created using natural methods—but were faster and safer to develop.

With continued refinement, this AI-powered approach could make personalized cancer treatments more available, more precise, and much faster to deliver. The future of immunotherapy is no longer just about understanding biology—it’s also about designing solutions with digital tools.

Note: The article above provided above by The Brighter Side of News.

Like these kind of feel good stories? Get The Brighter Side of News' newsletter.