Plasma Metabolomics Classifier Detects Glioma via Seven-Metabolite Liquid Biopsy Panel
Multi-omics profiling of 189 glioma tissue samples and 430 plasma samples identified aberrant alanine/aspartate/glutamate metabolism and TCA cycle signatures as universal features across glioma subtypes, detectable in plasma. A seven-metabolite liquid biopsy model achieved AUC of 0.964 for adult glioma and 0.925 for paediatric brain tumours in independent test sets, with higher sensitivity for glioma than pancreatic cancer, supporting tumour selectivity. A promising non-invasive diagnostic tool for a cancer type with limited liquid biopsy options.
The original study
Integrative Multi-Omics Analysis Reveals the Characteristic Metabolic Signature of Glioma and Enables Plasma-Based Liquid Biopsy.
- Authors
- Jiang Y, Lan Y, Wang Y, Chen S, Shen Y, Chu S, et al.
- Journal
- Research (Washington, D.C.)
- PMID
- 41878633
Original abstract
Liquid biopsy strategies for glioma leveraging metabolic features remain inadequately investigated. Herein, we performed liquid chromatography-mass spectrometry-based metabolomic and proteomic analyses on 189 tissue samples from 122 adult glioma patients, and nuclear magnetic resonance-based targeted metabolomic profiling on plasma samples from 430 participants encompassing 82 adult glioma patients, 53 pediatric primary brain tumor patients, 80 pancreatic cancer patients, and 215 nontumor controls. The results demonstrate that aberrations in "Alanine, aspartate, and glutamate metabolism" and "tricarboxylic acid (TCA) cycle" pathways are ubiquitous across subtypes and progression of glioma. Notably, these signatures could be captured in plasma, thereby reflecting shared metabolic features between tumor tissues and circulation. Based on these findings, we developed a liquid biopsy model comprising 7 plasma metabolites (including creatine, lactic acid, succinic acid, N,N-dimethylglycine, 2-oxoglutaric acid, acetic acid, and glutamic acid). This model achieved high diagnostic accuracy in independent test sets (area under the curve = 0.964 for adult glioma set; and 0.925 for pediatric primary brain tumor set). Meanwhile, the model exhibited a higher sensitivity of 0.885 for glioma compared to 0.800 for pancreatic cancer, providing evidence to support the tumor selectivity of the model. Together, we present a plasma-based metabolomic classifier that faithfully mirrors the core metabolic reprogramming of glioma and can serve as a readily available liquid biopsy tool.