Network-guided identification of candidate biomarkers for COPD-associated comorbidities using deep learning and real-world transcriptomic data

A COPD-centered network search plus deep-learning discrimination identified a five-protein exploratory biomarker panel—GTF2I, UBXN7, LIFR, SRGN, and FGF4—linked to specific COPD comorbidities.

• Why it matters: COPD comorbidity markers derived only from healthy controls may miss molecular changes from established COPD to a particular comorbidity, limiting comorbidity-specific biomarker discovery.
• What they did: The Human Protein Network (17,302 proteins; 492,195 relationships) was encoded with 20 graph features, and a functional-distance model ensemble trained on drug-target-indication links prioritized 224 COPD-nearby candidates (FDM score > 0.50). Candidate expression was then evaluated in independently assembled public COPD patient transcriptomic datasets using one-feature classifiers with cross-validation.
• The result: In cross-validation, balanced accuracy ranged from 0.838 to 0.970 and MCC from 0.726 to 0.941, yielding GTF2I for bacterial infection, UBXN7 for type 2 diabetes, LIFR for hypertension, SRGN for lung neoplasm, and FGF4 for pneumonitis. SRGN has direct experimental support in non-small-cell lung cancer, while the others remain hypothesis-level.

AI summary of the abstract · may contain errors · not medical advice · read the paper.