P05.05.A INTEGRATED MULTI-OMICS AND MULTI-SPECTRAL PROFILING OF PLASMA EXTRACELLULAR VESICLES REVEALS HIGHLY ACCURATE LIQUID BIOPSY BIOMARKERS FOR GLIOMA DIAGNOSIS.

Abstract

BACKGROUND Liquid biopsy approaches are revolutionising the management of patients with extracranial solid tumours. However, the techniques investigated have not been effective in neuro-oncology. There is therefore an urgent need to develop an accurate and reproducible liquid biopsy technique for neuro-oncology patients. Plasma small extracellular vesicles (sEVs) are proposed as an alternative liquid biopsy analyte for glioma. This study aimed to define, and validate, a comprehensive multi-omics/multi-spectral plasma sEV biomarker signature for glioma diagnosis. MATERIAL AND METHODS sEVs were separated from 1 mL plasma samples using our optimised size exclusion chromatography protocol. Following characterisation according to MISEV2023 guidelines, sEVs were comprehensively analysed using mass spectrometry proteomics, microRNA transcriptomics, attenuated total reflection Fourier transform infrared (ATR-FTIR) spectroscopy, and Raman spectroscopy. Machine learning approaches (Random Forest, K-Nearest Neighbours, Extreme Gradient Boosting) were used to identify the most effective signature that differentiated glioma patients’ samples from gender- and age-matched healthy volunteers’ samples. External validation using two independent patient cohorts and orthogonal multi-omics analysis is ongoing and will be presented. RESULTS Compared to healthy volunteers (n=48), glioma patients’ sEVs (n=56) were larger (p<0.001), with a greater concentration of >80 nm subpopulations (p=0.0024). sEVs -omics analysis identified 305 differentially abundant proteins, including coagulation and cholesterol metabolism associated proteins, and 77 differentially expressed microRNAs, including those associated with gliomas. ATR-FTIR and Raman spectral analysis demonstrated clear differences in overall spectral signature, with differences in sEVs protein, nucleic acid, and lipid composition. Machine learning approaches generated signatures of sEV protein, microRNA, ATR-FTIR, and Raman spectrums with excellent discriminating ability (AUCs 0.916-0.988), whilst the comprehensive integrated sEV biomarker signature could distinguish glioma patients’ samples from healthy volunteers’ samples with near-perfect accuracy (AUC 0.984). CONCLUSION Our results demonstrate that the machine learning integration of sEV multi-omics/multi-spectral datasets can achieve high-fidelity glioma detection, demonstrating clinical scalability for pre-operative diagnosis from 1mL plasma samples. This platform establishes a foundation for liquid biopsy-driven neuro-oncology frameworks.