Project Summary/Abstract Asthma is a chronic respiratory disease affecting about 340 million people worldwide, yet its causal biology, environmental risks, key cell types, and optimal treatments remain under-characterized. This difficulty is partly due to clinical heterogeneity, as different risk factors drive asthma for different people. Asthma subtype studies have already begun to reveal important aspects of this heterogeneity. However, asthma subtypes remain nascent and ambiguous and have not yet realized their potential utility for scientific studies and precision treatments. In particular, genetics has not been fully exploited for asthma subtyping, though it has a unique ability to assess the causal biological significance of subtypes and can identify key cell types; conversely, prior subtyping studies are susceptible to coincidental subtypes that are not directly relevant to asthma biology. Furthermore, prior studies have used basic methods which are liable to bias and low power. To address these limitations, we will develop a powerful and robust framework to pinpoint and genetically characterize asthma subtypes, and we will broadly apply it in large, deeply phenotyped, and diverse cohorts. Our study will identify novel subtypes and their demographic, genomic, cellular, and clinical etiologies, which can suggest precision treatments and improve power and interpretation in basic and translational research. Our work will improve genetic prediction of asthma, particularly in understudied populations. Finally, our approach and freely released methods will provide a broad template for complex trait subtyping. To accomplish these goals, we will study four large biobanks, which offer unprecedented sample size, clinical depth, and demographic diversity. We will use functional genomics to link genetic heterogeneity to causal and cell type-specific molecular mechanisms. We will build on our prior machine learning tools to identify subtypes, quantify their genetic and clinical significance, and infer their dominant cell types. Our methods are unique by correcting for confounding population structure, which is crucial for genetic subtyping: spurious genetic associations led prior studies to propose severely biased and regressive nosology. A key goal of this proposal is the PI’s retraining in asthma biology, pulmonology, and functional genomics. This will be achieved by close mentorship from Professors Carole Ober, Julian Solway, Yoav Gilad, and Matthew Stephens, as well as didactic courses in the UChicago Department of Medicine and Institute for Translational Medicine. This retraining will maximize the biomedical impact of our study by enabling the PI to deeply connect quantitative results to core facets of asthma pathology and will establish the PI as an independent asthma researcher who can optimally apply his statistics and machine learning background to tackle essential biomedical hurdles.