New insight into allergies

Artificial intelligence has uncovered previously hidden allergens within the trillions of microbes living in the human body, offering new insight into why some people develop allergies while others do not.

In research published in Cell Systems this month, scientists used a deep learning model to scan millions of proteins from microbes found in the gut and mouth. The model identified hundreds of proteins that could trigger allergic reactions, giving researchers a powerful new way to identify previously unknown allergens and better understand the biological mechanisms that drive allergic disease. This includes how differences in the proteins produced by an individual’s microbiome may influence their risk of developing allergies.

The study was led by Dr Kumar Thurimella, who carried out the research during his PhD with the Bioelectronics Systems Technology (BEST) research group in the Department of Chemical Engineering and Biotechnology (CEB). He is one of two lead authors of the study, supervised by Professor Róisín Owens and Dr Sergio Bacallado in the Department of Pure Mathematics and Mathematical Statistics, in collaboration with colleagues at Harvard Medical School and Massachusetts General Hospital.

Allergies affect hundreds of millions of people worldwide and are typically linked to external triggers such as pollen, foods, or dust mites. Microbes living naturally in the human body – known collectively as the microbiome – also produce proteins that interact with the immune system, but understanding their role in triggering allergic responses has been extremely limited until now.

To investigate, researchers developed the deep learning model, which identified hundreds of candidate allergens. When two of the top predictions were tested experimentally, both were confirmed to trigger allergic immune responses.

The AI was trained exclusively on serine proteases but also successfully identified a cysteine protease allergen, a different enzyme class with similar evolutionary patterns but distinct activity. It highlights the model’s ability to generalise and recognise underlying biological features associated with allergenicity.

Professor Róisín Owens said: “Allergic diseases are complex and, in many cases, we still do not fully understand what triggers them. This work shows how artificial intelligence can help us uncover previously hidden allergenic proteins within the human microbiome, providing new insight into how microbes interact with the immune system.

“It also provides a foundation for understanding how differences in the human microbiome may influence an individual’s susceptibility to allergic disease.”

The work also highlights the growing role of AI in biomedical discovery, particularly in uncovering hidden patterns within large and complex biological datasets, and demonstrates how interdisciplinary collaboration can boost understanding of human disease.

Dr Thurimella, now an MD candidate at the University of Colorado School of Medicine, completed the research as part of his PhD at Cambridge, supported by the Gates Cambridge Scholarship.