BIOM-39. PLASMA PROTEOMICS REVEAL KEY PRELIMINARY BIOMARKERS FOR GLIOBLASTOMA PROGRESSION AND DIFFERENTIAL DIAGNOSIS

Résumé fourni par la source

Abstract Adult-type diffuse gliomas, including glioblastoma (GBM), astrocytoma, and oligodendroglioma, are among the most aggressive brain tumors, often exhibiting poor prognosis. Diagnosis currently relies on MRI and biopsy, but for individuals ineligible for resection, non-invasive diagnostic tools are lacking. Additionally, distinguishing primary from recurrent disease and monitoring therapy response remain critical unmet needs. We evaluated seven candidate proteomic markers in plasma from individuals with GBM (n=143), astrocytoma (n=50), oligodendroglioma (n=32), and non-tumor controls (n=30) using research-use-only electrochemiluminescence assays (Meso Scale Discovery). Key questions included: (1) Do protein levels change between primary and recurrent GBM? (2) Do protein concentrations correlate with survival in primary vs. recurrent disease? (3) Can markers predict GBM tumor burden? (4) Can they distinguish GBM from other gliomas and controls? The results showed that neurofilament light chain (NEFL) increased significantly from primary to recurrent GBM (unadjusted p = 0.008). In primary GBM, low fatty acid binding protein 4 (FABP4) correlated with survival (adjusted p = 0.028), but no significance was observed in recurrent disease after adjustment. Machine learning (Lasso regression) showed poor performance in predicting tumor volume (cross-validated R² = 0.192). However, dimensionality reduction (PCA-UMAP) revealed distinct clustering of GBM samples versus astrocytomas, oligodendrogliomas, and controls. A random forest classifier trained on a 70:30 split achieved strong diagnostic performance (test AUC = 0.96, sensitivity = 0.95, specificity = 0.83). To conclude, plasma proteomic markers, particularly NEFL and FABP4, show promise for monitoring recurrence and prognostic stratification in GBM. While tumor volume prediction was unreliable, machine learning models excelled in differentiating GBM from other gliomas and controls, suggesting clinical utility for non-invasive diagnosis. Further validation is warranted to refine biomarker panels for precision oncology applications.

Ce résumé expose les affirmations des auteurs. BNTIC ne l’interprète pas comme une validation indépendante des résultats.