One AI model both predicts T cell recognition and designs new receptors.

Your immune system uses T cells to spot threats like viruses and tumors. Special sensors called T cell receptors grab onto tiny pieces of target molecules. Usually, computer tools either predict what a receptor can bind or try to design new ones, but not both. Researchers built an artificial intelligence model called OmniTCR to tackle both jobs together. It learned by reading millions of human immune sequence records, training on both single receptor parts and complete binding pairs. Because it learned these combined rules, OmniTCR accurately predicts whether receptors will bind to unfamiliar targets. It also spots the difference between healthy immune samples and cancer samples, and it designs realistic new receptor sequences to match specific targets.

Why it matters

T cells are the immune system's search-and-destroy cells, and they recognise threats through receptor proteins. Until now, predicting which receptors bind what and designing new ones were separate tasks, so data stayed fragmented. OmniTCR unifies them, which could make personalised cancer immunotherapy and vaccine design faster and more reliable.

Questions answered