A new AI system is changing how we approach cancer immunotherapy, shrinking the treatment timeline from years to weeks.
WASHINGTON, D.C., United States – August 4, 2025: This new AI platform speeds up the creation of custom cancer immunotherapies. It helps scientists create immune cells that can precisely target and wipe out tumors much faster than before. The method was written about in the journal Science. It uses a system of AI models to make special proteins that act like molecular keys for a patient’s T cells. These keys guide the T cells to find and attack cancer cells. It used to take years of hard work to do this, but now it can be done in just four to six weeks. This could really change how we treat cancer in the future. The researchers at the Technical University of Denmark (DTU) who led this project are excited because this could mean that personalized treatments for different kinds of cancers become easily available to people.
Creating New Cancer-Fighting Cells
The most important thing about this new tech is that it can create proteins from scratch. These proteins are designed to connect to molecules related to cancer. These molecules, called pMHCs, are parts of proteins that show up on the surface of cancer cells. T cells can recognize these pMHCs. The old ways of finding T-cell receptors that can target these pMHCs naturally are slow and don’t always work.
The AI platform solves this problem by inventing completely new proteins, called minibinders. These AI-made minibinders are built to fit the pMHCs on cancer cells, like a key fitting into a lock. When these are put into a patient’s T cells, they give the immune system a new and specific way to find the tumor.
Using Several AI Models
This innovation isn’t just one AI doing all the work. It’s a smart system of three connected models. First, there’s a generative AI model called RFdiffusion. It comes up with many protein shapes that could connect to the cancer target. This model is trained with lots of data about known protein structures, so it can quickly make thousands of possible designs.
Then, a second model steps in to make the designs better. It figures out the exact sequences of amino acids needed for a successful protein. After that, a third AI model does a virtual safety check. It compares the potential proteins against a database of pMHCs found on healthy cells. This is very important because it makes sure the new minibinders won’t attack the patient’s healthy tissues by mistake. This lowers the risk of bad side effects.
Inspired by Current Treatments
This new method is based on how CAR T-cell therapy works. CAR T-cell therapy is a type of immunotherapy where a patient’s T cells are changed genetically to show a Chimeric Antigen Receptor (CAR). CAR T-cells have been successful in treating some blood cancers, like lymphoma and leukemia. But it’s been tougher to use them for solid tumors.
The AI-made minibinders provide a new way to handle these challenges. The treatment would be similar to CAR T-cell therapy. A patient’s blood is taken, and their immune cells are changed in a lab with the AI-designed proteins. Then, these enhanced cells, called IMPAC-T cells by the researchers, are put back into the patient’s body to find and destroy the cancer.
Tested on a Known Cancer Target
To be sure the platform works, the research team tried it out on NY-ESO-1. This is a protein that’s commonly found in many types of cancers. The lab results showed that the AI-designed proteins were able to guide T cells to kill melanoma cells and stop tumor growth. This shows that the platform can be used to design targeted treatments for any cancer that has a unique marker on its surface.
The study was headed by Associate Professor Timothy P. Jenkins from the Technical University of Denmark. He says it represents a big step forward. The work is based on computer protein design technologies that won the 2024 Nobel Prize in Chemistry. This shows how quickly this scientific area is improving.
From the Lab to the Clinic: What’s Next?
Even though this research is a major step, it will still take several years before it can be used in clinics. Right now, the study is just a proof-of-concept. This means that more research and lots of clinical trials are needed to test if it’s safe and works well in people.
Still, the speed of the AI process changes everything. Being able to design a new treatment option in just a few weeks greatly cuts down the time needed for preclinical work. This faster pace could really decrease the amount of time before new immunotherapies are available to patients.
A New Type of Personalized Medicine
This technology could have huge effects. It could bring about a new kind of personalized cancer treatment. Treatments could be custom-made for each patient’s specific tumor. This would improve treatment results and lessen the bad side effects that often come with treatments like chemotherapy and radiation.
This is a great example of how artificial intelligence is becoming more and more important in scientific discovery and medicine. As AI models get better, they could invent a whole new group of treatments, not just for cancer but also for many other illnesses.
