Related Experiment Videos
Noninvasive early detection and grading of pneumoconiosis via plasma proteomics and machine learning: PRSS3 as a
Yizhuo Tian1, Mingyao Wang1,2, Zhifei Hou3,4
1State Key Laboratory of Common Mechanism Research for Major Diseases, Department of Biochemistry and Molecular Biology, Institute of Basic Medical Sciences, Institute of Pathogen Biology, Chinese Academy of Medical Sciences & Peking Union Medical College, Beijing, 100005, China.
Background:
Coal-dust, a persistent airborne pollutant, induces dose-related pulmonary fibrosis; however, plasma biomarkers for pre-clinical toxicity remain lacking.
Methods:
We enrolled 158 participants, including 28 healthy controls (HCs), 30 dust-exposed workers (DEWs), and 100 patients with coal workers' pneumoconiosis (CWP) at different stages (n CWP-I=40, n CWP-II=30, n CWP-III=30). Plasma proteomic profiling was performed via data-independent acquisition (DIA) mass spectrometry. Differentially expressed proteins were identified and functionally annotated. Key proteins were selected and multiple machine learning algorithms were employed to construct and validate predictive models.
Results:
We identified 1,239 plasma proteins, including 645 high-confidence candidates. Functional enrichment revealed significant associations between disease progression and pathways such as PPAR signaling, cholesterol metabolism, Epstein-Barr virus infection, and the pentose phosphate pathway. These alterations converge on dysregulated lipid metabolism, chronic inflammatory signaling and virus-induced immune evasion, suggesting a metabolic-immune axis that orchestrates early fibrotic progression. We successfully constructed the first plasma proteomics-based machine learning models for pneumoconiosis grading and early screening. Notably, a single biomarker, PRSS3, demonstrated exceptional performance in distinguishing DEW patients from early-stage pneumoconiosis patients (CWP-I), achieving an area under the curve (AUC) of 1.00 and an accuracy of 1.00 in the training set and an AUC of 1.00 with an accuracy between 0.93 and 1.00 in the validation set.
Conclusion:
This study establishes innovative machine learning-based models for the grading and early screening of pneumoconiosis via plasma proteomics. The identification of PRSS3 as a potential biomarker highlights the clinical utility of our approach. These findings provide a foundation for noninvasive diagnostic strategies and future translational research in occupational lung diseases.
