Using AI to Find Better Tumour Targets for CAR T-Cell Therapy

CAR T-cell therapy has transformed the treatment of some blood cancers, but its use in solid tumours, such as colorectal cancer, has been much more challenging. One of the biggest obstacles is finding targets that are found on cancer cells but not on healthy cells, so the treatment can attack the cancer safely. 

What is CAR T-cell therapy? 

CAR T-cell therapy is a type of immunotherapy that uses a person’s own immune cells to fight cancer. Doctors collect a patient’s T cells (a type of white blood cell), genetically modify them in a laboratory so they can recognize a specific protein on cancer cells, and then return them to the patient’s body. Once infused, these engineered CAR T cells can find, attack, and destroy cancer cells. 

Traditionally, identifying these targets has been a slow, complex process that requires researchers to analyze large amounts of biological data by hand. In a recent study, researchers developed a new approach that uses artificial intelligence (AI), including large language models (LLMs), to speed up the search for promising CAR T-cell targets. 

The AI analyzed genetic data from cancer and healthy tissues and combined this with information from public databases to identify targets that are: 

• Found mainly on cancer cells rather than healthy cells.  

• Located on the cell surface, making them accessible to CAR T cells.  

• Likely to be safe and practical for developing new treatments.  

Using this approach, the researchers identified GPNMB as the most promising target. They then developed a CAR T-cell therapy directed against GPNMB and found that it showed strong anti-tumor activity in laboratory models of colorectal cancer, melanoma, and leukemia. 

Take-home message 

 This AI-powered strategy could help researchers discover new CAR T-cell targets more quickly and efficiently, potentially leading to new immunotherapy options for a range of cancers, including colorectal cancer. While the findings are promising, the research is still in the preclinical stage and more studies are needed before this approach can be tested in patients.