Serum Metabolomics Profiling Coupled with Machine Learning Identifies Potential Diagnostic and Prognostic Candidate Markers in Meningioma Using Raman Spectroscopy, ATR-FTIR, and LC-MS/MS.

Abstract

Meningioma, the most prevalent brain tumor, poses significant challenges due to its unclear transition from low-grade to aggressive forms, with limited knowledge about grade-specific markers. We have utilized vibrational spectroscopic techniques such as ATR-FTIR and Raman spectroscopy, alongside LC-MS/MS-based mass spectrometry to understand the systemic cues and evaluate them for clinical practice. The acquired Raman and ATR-FTIR spectra of 46 meningioma patients (27 low-grade and 19 high-grade) and 8 healthy individuals revealed 98.15% and 83.33% accuracy based on PC-LDA. The grade classification revealed an accuracy of around 70%, implying the presence of subtypes and transition phases. The observed alterations corresponded to lipids, nucleic acids, and proteins. Further, the LC-MS/MS-based study identified different derivatives of cholines, indoles, lipids, sphingosine, tryptophan, and their respective metabolic pathways as contributors in tumorigenesis and progression. Further, PRM-based targeted validation and feature selection was carried out on 43 meningioma patients and 17 healthy controls. Glycochenodeoxycholic acid, indole-3-acetic acid, trans-3-indoleacrylic acid, glycodeoxycholic acid, 5α-dihydrotestosteroneglucornide, and glycocholic acid segregated meningioma samples with an accuracy of around 90% while features like indole-3-acetic acid, stercobilin, sphingosine-1-phosphate, deoxycholic acid, and citric acid could classify grades with around 70% accuracy. These findings suggest that further validation across larger cohorts could enhance its usage in clinical settings.