血漿由来エクソソームを用いた肺がんスクリーニング：液滴コーティング堆積ラマンスペクトルと機械学習による評価

Lung cancer is the leading cause of cancer-related mortality worldwide, making early screening crucial for improving patient survival. In recent years, exosomes have garnered significant attention as promising biomarkers for the detection of lung cancer. Their easy isolation from body fluids, such as blood and urine, makes them a perfect sample for liquid biopsy, while liquid biopsy is widely used in clinical research. Droplet coating deposition Raman (DCDR) spectroscopy is well-suited for exosome detection due to its molecular fingerprints, non-invasiveness, low sample volume requirements, and minimal/no sample preparation. In this study, we combined DCDR technology with machine learning algorithms to screen for lung cancer based on plasma-derived exosomes. High-quality exosomes were isolated from clinical blood samples via ultracentrifugation, exhibiting a characteristic cup-shaped morphology with an average diameter of 130 nm and expressing canonical exosome markers (CD63 and CD81). Although subtle differences were observed between the Raman spectra of exosomes from lung cancer patients and from healthy individuals, principal component analysis (PCA) revealed the presence of a batch effect across the samples. To enable diagnosis while minimizing the impact of batch effect, the support vector machine (SVM) model outperformed the linear discriminant analysis (LDA) model, achieving 95.60% accuracy (area under the curve (AUC) = 0.996) at the spectrum level and 100% accuracy at the patient level. These results demonstrate that Raman spectroscopy is an up-and-coming tool for rapid lung cancer screening, offering the advantages of low cost, ease of operation, low sample volume requirements, and high speed.