AI-based prediction of aspirin-exacerbated respiratory disease using nasal epithelial mRNA expression profiles.

BACKGROUND: Aspirin-exacerbated respiratory disease (AERD) is a distinct asthma endotype marked by asthma, nasal polyposis, and respiratory reactions to COX-1 inhibitors. Early and accurate identification of AERD remains clinically challenging. OBJECTIVE: We sought to develop and externally validate an artificial intelligence (AI)-based diagnostic model that uses nasal epithelial mRNA expression profiles to accurately identify AERD. METHODS: mRNA gene expression profiles were obtained from nasal epithelial brushing in 71 subjects with AERD and 57 without AERD. AI models were trained to predict an AERD diagnosis in a training cohort using gene expression alone, which was then validated on an independent validation cohort. RESULTS: The clinical data analysis revealed noteworthy findings of AERD: 29% reported cutaneous manifestations during nonsteroidal anti-inflammatory drug reactions, 50% experienced symptoms related to alcohol consumption, and 59% required 2 or more sinus surgeries. AERD was predicted with an accuracy of 93% in the training cohort and 83% in the independent validation cohort. The top AERD-predicting genes included IL1RL1 (IL-33 receptor) and CLC (Charcot-Leyden crystal protein), which are known to be important to AERD pathogenesis. CONCLUSIONS: Nasal transcriptomics can predict AERD diagnosis accurately and may improve disease understanding, enabling earlier and more precise endotype-based diagnosis and management.