Related Experiment Video
Updated: Nov 29, 2025
MicroRNA Based Liquid Biopsy: The Experience of the Plasma miRNA Signature Classifier MSC for Lung Cancer Screening
Published on: October 26, 2017
Early lung cancer diagnostic biomarker discovery by machine learning methods
Ying Xie1, Wei-Yu Meng1, Run-Ze Li1
1State Key Laboratory of Quality Research in Chinese Medicine/Macau Institute for Applied Research in Medicine and Health, Macau University of Science and Technology, Macau (SAR), China.
This study identifies six plasma metabolic biomarkers for early lung cancer detection in Chinese patients using metabolomics and machine learning. These biomarkers show high accuracy in distinguishing early-stage lung cancer from healthy individuals, supporting blood-based screening.
Area of Science:
- Oncology
- Biochemistry
- Computational Biology
Background:
- Early diagnosis significantly improves lung cancer patient survival rates.
- Blood-based screening offers potential for increased early detection uptake.
- Identifying reliable plasma biomarkers is crucial for non-invasive lung cancer diagnosis.
Purpose of the Study:
- To discover plasma metabolic biomarkers for early lung cancer diagnosis in Chinese patients.
- To apply an interdisciplinary approach combining metabolomics and machine learning for biomarker discovery.
- To evaluate the diagnostic performance of identified biomarkers for stage I lung cancer.
Main Methods:
- Targeted metabolomic analysis of 61 plasma metabolites using LC-MS/MS.
- Application of machine learning algorithms, including FCBF and Naïve Bayes, for biomarker selection and prediction.
- Study cohort comprised 110 lung cancer patients and 43 healthy individuals.
Main Results:
- A combination of six metabolic biomarkers achieved high discrimination between stage I lung cancer and healthy individuals (AUC=0.989, Sensitivity=98.1%, Specificity=100.0%).
- The top 5 metabolic biomarkers identified by the FCBF algorithm show potential for early lung cancer screening.
- Naïve Bayes classifier demonstrated effectiveness for early lung tumor prediction.
Conclusions:
- The identified plasma metabolic biomarkers are promising for the early, accurate, and non-invasive diagnosis of lung cancer.
- The interdisciplinary approach combining metabolomics and machine learning is feasible and effective for lung cancer biomarker discovery.
- This methodology holds potential for adaptation to other cancer types beyond lung cancer.
