Multi-omics Insights on the Disease Trajectory Linking COPD and Lung Cancer

Chronic obstructive pulmonary disease (COPD) and lung cancer (LC) share tobacco exposure and chronic pulmonary inflammation as principal drivers, but the molecular architecture of the trajectory linking the two has not been delineated in a single integrative framework. Drawing on 483,728 UK Biobank participants without baseline COPD or LC, we constructed a four-state, nine-transition multistate model spanning health, single COPD, single LC, COPD–LC and LC–COPD multimorbidity, and death, and integrated three baseline omics layers: genomics (polygenic risk scores; N = 469,072), nuclear magnetic resonance metabolomics (251 markers; N = 465,676) and Olink Explore plasma proteomics (2,922 proteins; N = 50.239). Over a median follow-up of 14 years, 22,376 participants developed COPD, 4,908 developed LC, and 1,428 progressed to multimorbidity. Participants with COPD–LC carried the worst prognosis, with a restricted mean survival of 3.0 years. Across the nine transitions, distinct and shared multi-omics signatures were identified, with metabolomics carrying a higher proportion of overlapping signal and proteomics carrying a higher proportion of transition-specific signal; pathway analyses converged on cytokine–cytokine receptor interaction, PI3K–Akt, NF-κB and MAPK signalling. A LASSO-derived proteomic risk score (Prot-Score) outperformed genomic and metabolomic scores across all nine transitions, with ΔC-statistic improvements of 0.06–0.35 over a covariate-only base model and 10-year AUCs of 0.68–0.96 in the combined cohort (N = 22,789). Smoking-stratified analyses and marker-by-smoking interaction tests indicated that the principal associations and the predictive advantage of Prot-Score were not driven by residual confounding from tobacco exposure. These findings provide an integrative molecular and prognostic characterisation of the COPD–LC trajectory and identify plasma proteomics as the omics layer carrying the greatest incremental prognostic information; external validation in independent cohorts is required before any clinical application can be considered.