Abstract
Background: Allergen multiplex assays are increasingly used as a precision medicine approach in difficult-to-diagnose allergic patients. It requires extensive knowledge in molecular allergology and appears very time-consuming for interpretation. We hypothesized that a nationwide dataset able to support artificial intelligence-assisted allergy diagnosis may improve the management of allergic patients.Method: The French Society of Allergology (SFA) and the Health Data Hub (HDH) partnered for the development of a retrospective dataset. Allergen multiplex collection was led by the specialized AllergoBioNet network of clinical laboratories. Board-certified allergists assessed allergy diagnosis, clinical history, and therapeutic management. Data scientists, epidemiologists and public health specialists from the Desbrest Institute of Epidemiology and Public Health (IDESP) and Trustii, encoded clinical items as semantic attributes and supervised the anonymization in compliance with European regulation 2016/679 (General Data Protection Regulation, GDPR) and French data protection laws.Results: Data were collected from 15 university hospitals spanning the French territory. A wide panel of complex conditions was obtained, including food and airborne allergy and anaphylaxis in 4000 patients aged 0–80 years. In a subset of patients, images from processed allergen multiplexes were collected as raw data for IgE antibody quantitation. The dataset will be open following an international crowdsourced machine learning competition helding from April 1st to May 31st 2023.Conclusion: We report on the methodology and establishment of the first nationwide dataset of allergen multiplex and associated diagnostic and therapeutic data representative of allergies encountered in a Western European country. This dataset paves the way for an open-source diagnostic prediction tool for the practicing allergist.