Asthma has long been recognized as a heterogeneous, complex syndrome, both clinically and pathogenetically. Identifying asthma endotypes and defining genetic and environmental contributions have been challenging. We reasoned that individuals with asthma plus different comorbidities (e.g., cardiovascular vs gastrointestinal diseases) may represent distinct endotypes of asthma that arise in disparate genetic variation and life-time environmental exposure backgrounds. To test this hypothesis, we first computationally discovered 22 distinct asthma comorbidity patterns (“asthma comorbidity subgroups”) using diagnosis records for over 151 million US residents. Our model-based inference defined each subgroup by its frequency distribution of comorbid diseases. Eleven of the 22 subgroups could also be found in the UK Biobank; we then assigned individuals there to one of the eleven subgroups, and conducted genome-wide association analyses by comparing asthma cases and non-asthma controls within each subgroup as well as in all subgroups combined. We identified 109 independent loci significantly associated with asthma, of which 52 were replicated in a follow up multi-ancestry meta-analysis across different ethnicity subsamples taken from UK Biobank, US BioVU, and Biobank Japan. In particular, 14 loci conferred asthma risk in multiple subgroups as well as in all these subgroups combined; importantly, another six loci conferred asthma risk in one subgroup only. Furthermore, we observed that the strength of association between asthma and each of 44 health-related phenotypes varied dramatically across subgroups; for example, red blood cell production and accumulation was especially associated with asthma in the subgroup marked by high frequencies of lung diseases, while strong association with local environment and physical activities was observed in the subgroup marked predominantly by diabetes comorbidity. This work reveals subpopulations of asthma patients that are distinguished by comorbidity patterns, asthma risk loci, gene expression, and health-related phenotypes, and can therefore be considered as different asthma endotypes. The presented workflow has general applicability to characterize endotypes of other heterogeneous complex diseases.

electronic health record;comorbidity pattern;clustering;disease subtyping