Plasma proteomic profiling with machine learning identifies immunometabolic perturbations associated with smoking and smoking cessation.

Cigarette smoking is the leading modifiable risk factor for chronic obstructive pulmonary disease (COPD) and is associated with systemic molecular alterations linking exposure to disease. We profiled the plasma proteome (2,920 proteins) of 38,133 UK Biobank participants and applied machine-learning approaches to characterize smoking-associated alterations and develop risk models. The circulating proteome exhibited dose-dependent changes, with heavy smoking associated with prominent immunometabolic and inflammatory signatures. An 11-protein panel discriminated COPD with an AUC of 0.844, outperforming models based on smoking history alone and distinguishing individuals with higher versus lower susceptibility. Model-based analyses further highlighted a subgroup of heavy smokers-characterized by older age, higher body mass index, and leptin-skewed inflammatory profiles-showing molecular features associated with elevated COPD risk after smoking cessation. Together, these findings characterize systemic proteomic patterns associated with smoking exposure and support plasma-based stratification of COPD risk.