Researchers have developed a powerful new artificial intelligence tool, ImmunoMatch, that can predict which antibody components naturally pair together. This advance can significantly accelerate antibody-based drug development and deepen our understanding of how the immune system works.
Antibodies, the proteins our bodies use to detect and neutralize pathogens, are built from two parts: a heavy chain and a light chain. Each B cell produces a unique combination, but only certain pairs form stable, functional antibodies. Until now, deciphering which heavy and light chain sequences naturally pair, or form stable pairs when engineered, has remained a major challenge. Challenging the belief that pairing occurs in a random fashion, the researchers used Artificial Intelligence on millions of paired heavy and light chains to show that their pairing is predictable. The resulting computational tool, ImmunoMatch, offers unprecedented speed and accuracy in identifying correct pairings.
Using thousands of real antibody sequences from healthy human donors, the UCL-led research team trained a machine-learning model to recognise subtle patterns that distinguish true biological heavy–light chain pairs from random combinations. The resulting system can score how likely any two chains are to form a compatible antibody.
The tool performed strongly across training and independent datasets, outperforming traditional approaches that rely on simple gene usage or antibody hypervariable regions. Importantly, it also proved effective on spatial VDJ sequencing data, where heavy and light chains are measured in tissue samples but not captured as pairs. Using ImmunoMatch, researchers could reconstruct probable pairings from these complex datasets, opening new possibilities for studying immune responses within tumours, lymph nodes and other tissues.
The study also revealed a previously unrecognized biological trend: as B cells mature, their antibody chains show increasingly specific and optimized pairing. This suggests that part of the immune system’s refinement process involves fine-tuning which chains work best together.
For the first time, we can computationally infer which heavy and light chains truly belong together. ImmunoMatch paves the way to learn the rules governing how functional antibodies are assembled to overcome different immune challenges. It gives researchers a window into the hidden logic of antibody pairing and we believe it will greatly speed up the discovery and engineering of therapeutic antibodies.”
ImmunoMatch could help scientists more rapidly design therapeutic antibodies, prioritise stable chain combinations and explore the immune system in unprecedented detail using large-scale sequencing datasets.
Links
Research article in Nature Methods
UCL Research Profiles: