Integrative multi-omic analyses identify major axes of heterogeneity in chronic obstructive pulmonary disease and uncover their molecular contributors.

Chronic obstructive pulmonary disease (COPD) is a debilitating and progressive lung disease that affects millions of people worldwide. There is a continuing clinical need to characterize COPD at the molecular level to be able to identify the multi-omic biomarkers of its pathogenesis and to enable more accurate diagnoses and more effective treatment. We used Multi-Omics Factor Analysis (MOFA) to jointly analyze genomic, blood transcriptomic, and plasma proteomic data collected from 1,872 participants in the Genetic Epidemiology of COPD study who had moderate to very severe COPD. Five latent factors identified by MOFA were associated with COPD-related lung function, chest computed tomography (CT) imaging, and blood count phenotypes, as well as all-cause mortality. The top genetic, transcriptomic and proteomic contributors to these latent factors were also individually associated with COPD-related outcomes. Moreover, factor loadings and expression levels of top omic drivers helped distinguish between patient subgroups. Quantitative trait loci analysis of a latent factor that was jointly driven by transcriptomics and proteomics revealed potential common genetic control of gene expression and protein abundance. Polygenic risk scores derived from a genomics-driven latent factor were associated with chest CT imaging and lung function phenotypes, and these associations were replicated in an independent COPD cohort. Together, our results suggest the potential of integrative omic approaches to identify the major axes of heterogeneity in COPD and uncover the multi-omic interplay between the contributors to each axis.