Metabolomic Subtyping and Machine Learning-Based Diagnosis Reveal Clinical Heterogeneity in Silicosis

Area of Science:

• Occupational Health
• Metabolomics
• Pulmonary Medicine

Background:

• Silicosis is a global occupational hazard with varied disease progression.
• The metabolic underpinnings of silicosis heterogeneity are not well understood.

Purpose of the Study:

• To investigate metabolic mechanisms driving silicosis heterogeneity.
• To identify potential biomarkers for silicosis detection and subtyping.

Main Methods:

• Case-control study with 156 silicosis patients and 132 controls.
• Untargeted plasma metabolomics using liquid chromatography-mass spectrometry (LC-MS/MS).
• Non-negative matrix factorization (NMF) clustering, WGCNA, and machine learning were employed.

Main Results:

• 860 differentially abundant metabolites identified, including pathogen-associated compounds.
• Two distinct metabolic subtypes (NMF1, NMF2) revealed, with NMF2 showing higher pulmonary infection risk.
• NMF2 linked to inflammation and lipid peroxidation; NMF1 associated with fibrosis markers.
• A dual-metabolite panel (tyrosocholic acid, PI) achieved AUC > 0.85 for silicosis detection and subtyping.

Conclusions:

• Silicosis exhibits significant metabolic heterogeneity.
• Identified subtypes offer potential for improved patient stratification and targeted interventions.
• Metabolomic insights pave the way for enhanced early detection and treatment strategies.