Related Experiment Video
Updated: Aug 9, 2026
MicroRNA Based Liquid Biopsy: The Experience of the Plasma miRNA Signature Classifier (MSC) for Lung Cancer Screening
Published on: October 26, 2017
An explainable multi-label Diagnostic Prediction Model for lung diseases based on proteomic biomarkers
Yan Wang1, Jie Tan2, Xinjun Li3
1College of Medical Information and Artificial Intelligence, Shandong First Medical University and Shandong Academy of Medical Sciences, Binzhou People's Hospital Affiliated to Shandong First Medical University, Jinan, Shandong, China.
This study developed an explainable machine learning model using proteomic data to accurately differentiate between lung cancer, tuberculosis, and pneumonia. The model achieved high accuracy, offering a promising tool for complex lung disease diagnosis.
Area of Science:
- Pulmonary Medicine
- Bioinformatics
- Machine Learning
Background:
- Differential diagnosis of pneumonia, tuberculosis, and lung cancer is challenging due to overlapping symptoms and comorbidities.
- Traditional diagnostic methods have limitations in accurately distinguishing these complex lung diseases.
Purpose of the Study:
- To develop and validate an explainable machine learning model for multi-label classification of lung cancer, tuberculosis, and pneumonia using proteomic data.
- To overcome diagnostic limitations and improve the accuracy of differentiating these lung diseases.
Main Methods:
- Collected bronchoalveolar lavage fluid proteomic data from 358 patients.
- Constructed a voting ensemble model integrating XGBoost, Random Forest, and Gradient Boosting algorithms.
- Applied a 1.5-fold weighting strategy for the cancer class and evaluated using 5-fold cross-validation and an external cohort.
Main Results:
- The ensemble model achieved high performance with AUC values of 0.912 for tuberculosis and 0.813 for cancer detection.
- Demonstrated excellent performance, outperforming seven baseline models.
- Achieved 86% overall accuracy in the independent external validation cohort and identified key protein biomarkers.
Conclusions:
- An explainable, multi-label classification model based on proteomics was established for differential diagnosis of complex lung diseases.
- The model shows good performance and interpretability, offering potential for precise lung disease diagnosis.
- Provides a valuable reference for diagnosing lung diseases, particularly those with comorbidities.