Related Experiment Video
Updated: Aug 2, 2025
08:02
Visualization of Metabolites Identified in the Spatial Metabolome of Traditional Chinese Medicine Using DESI-MSI
Published on: December 16, 2022
1.9K
XAI-enabled neural network analysis of metabolite spatial distributions
Wenwu Ma1,2,3,4,5, Lanfang Luo2,3,4,5, Kun Liang2,3,4,5
1Department of Life Sciences and Medicine, University of Science and Technology of China, Hefei, China.
Analytical and Bioanalytical Chemistry
|April 21, 2023
Summary
Deep neural networks identified key metabolite biomarkers for distinguishing tissue types. This approach achieved over 98% accuracy in muscle aging and cancer studies, revealing propanoate metabolism as a potential anti-aging target.
Area of Science:
- Metabolomics
- Artificial Intelligence
- Biomarker Discovery
Background:
- Mass spectrometry imaging (MSI) generates high-dimensional data.
- Identifying tissue-specific metabolite biomarkers is challenging.
- Existing machine learning methods may require data preprocessing like dimension reduction.
Purpose of the Study:
- To develop and validate a deep learning model for analyzing MSI data.
- To utilize explainable artificial intelligence (XAI) for biomarker identification.
- To discover novel metabolite biomarkers for tissue classification and therapeutic targets.
Main Methods:
- Deep neural networks, including a custom deep convolutional neural network (Channel-ResNet10), were employed.
- Explainable artificial intelligence (XAI) methods, specifically channel selection, were used to identify key metabolite features.
- Metabolite biomarkers were analyzed using MetaboAnalyst for pathway enrichment.
Main Results:
- Channel-ResNet10 demonstrated superior classification accuracy (>98%) compared to seven other machine learning methods.
- The XAI method successfully identified critical metabolite features driving classification.
- In young vs. aged mouse muscle, differentially abundant metabolites were enriched in the propanoate metabolism pathway.
Conclusions:
- Deep learning models, particularly Channel-ResNet10, are highly effective for MSI data analysis.
- XAI facilitates the identification of crucial metabolite biomarkers.
- Propanoate metabolism is a potential therapeutic target for anti-aging interventions.
