Revealing metabolite biomarkers for acupuncture treatment by linear programming based feature selection
Yong Wang1, Qiao-Feng Wu, Chen Chen
1Academy of Mathematics and Systems Science, Chinese Academy of Sciences, Beijing. ywang@amss.ac.cn
BMC Systems Biology
|October 11, 2012
Summary
This study introduces a new method to find metabolite biomarkers for acupuncture's molecular effects. The Linear Programming based Feature Selection (LPFS) method identifies key molecules, aiding understanding of acupuncture's mechanism.
Area of Science:
- Metabolomics
- Systems Biology
- Molecular Mechanisms
Background:
- Acupuncture, a Traditional Chinese Medicine (TCM) practice, is increasingly used in Western countries.
- The molecular mechanisms underlying acupuncture's effects, particularly differences between acupoints, remain poorly understood.
- This knowledge gap hinders the broader acceptance and application of acupuncture.
Purpose of the Study:
- To elucidate the molecular mechanisms of acupuncture by identifying metabolite biomarkers.
- To develop and validate a novel computational method for biomarker discovery in high-dimensional metabolic data.
- To investigate potential differences in molecular effects across various acupoints.
Main Methods:
- Development of a Linear Programming based Feature Selection (LPFS) method for identifying metabolite biomarkers.
- Generation and analysis of high-throughput metabolic profiles from human acupuncture treatment at multiple acupoints.
- Utilizing an optimization model with nearest centroid to minimize selected features and cross-validation error.
- Comparison of LPFS performance against established methods like SVM-RFE and SMLR.
Main Results:
- The LPFS method successfully identified small sets of metabolites with low standard deviation and significant shifts, indicative of robust biomarkers.
- Several metabolite biomarkers associated with acupuncture treatment were discovered, providing candidates for further mechanistic studies.
- Comparative analysis of biomarkers from five specific acupoints (ST36, ST21, ST3, GB34, BL40) revealed similarities and differences, suggesting acupoint specificity.
Conclusions:
- Metabolic profiling is a promising approach for investigating the molecular mechanisms of acupuncture.
- The developed LPFS method demonstrates superior performance in selecting key molecules compared to existing techniques.
- LPFS is a versatile methodology applicable to other high-dimensional data analyses, including cancer genomics.
