用于NF1相关周围神经鞘瘤分型的循环蛋白质组学特征

Purpose : Early detection and interception of malignant transformation in Neurofibromatosis Type 1 (NF1) associated peripheral nerve sheath tumors are challenging due to insensitive clinical symptoms, limited specificity of standard of care imaging, and the limited negative predictive value of invasive tissue biopsy. Novel non-invasive and tumor site-agnostic surveillance assays provide potential for early diagnosis and intervention, especially during the critical transformation from benign plexiform neurofibromas (PN) to pre-malignant atypical neurofibromas (AN), defined by histopathology. We hypothesize that circulating proteins in the plasma accurately distinguish the spectrum of PN, AN, and malignant peripheral nerve sheath tumors (MPNST). Methods : 118 plasma samples (Healthy n = 10, PN n = 29, AN n = 25, and MPNST n = 54) from 79 patients seen at the NIH were analyzed using a proximity extension assay (PEA) panel for 1461 proteins ( Olink ). Unique protein signatures were identified using a one-versus-all comparison: PN-versus-all (PvA) with ANOVA and post-hoc Tukey honestly significant difference (HSD) of normalized protein expression (NPX) outputs from PEA. Significant proteins had an NPX difference ≥ 1.2 and p adj <0.05. Individual proteins' performances were assessed using Youden's index and a receiver operating characteristic (ROC) curve. Given PN's high prevalence in NF1 and the high specificity of the identified proteins, a PN-specific predictive signature was prioritized for downstream analysis due to its clinical utility. Individual PN-associated protein signatures were integrated using a Support Vector Machine Learning Model (SVM) with leave-one-out cross-validation (NPX difference ≥ 1.2 and p adj <0.05). Results: 114 proteins were significant for PvA comparisons. Individual protein's performance for PvA had a median AUC of 0.64 (IQR: 0.62-0.67), median sensitivity of 0.38 (IQR: 0.33-0.47), and median specificity of 0.97 (IQR: 0.86-1.0). SVM significantly improved the PN-signature's (AUC: 0.954),sensitivity (0.69, 20/29 PN), and specificity (0.99, one AN was misclassified as PN). The integrated PN protein signature predicts with high confidence whether a patient's tumor burden remains PN or has transformed to AN or MPNST (PPV: 0.95, NPV: 0.91). Conclusions : This pilot demonstrates that circulating proteomics non-invasively distinguish PN from pre-malignant and malignant disease states in NF1. The PN protein profile shows significant dysregulation of protein expression, protein-protein interactions, and biological pathways. Finally, PEA uses just 40 μL of plasma per sample, enabling the integration of non-invasive orthogonal biomarkers, such as cell-free DNA, from a single tube of blood.