Circularly Polarized SE-SERDS of Circulating cfDNA for Pretreatment Prediction of Nasopharyngeal Carcinoma Recurrence

Abstract Precise pretreatment prediction of nasopharyngeal carcinoma (NPC) recurrence remains a clinical challenge. This study develops a circularly polarized surface-enhanced shifted-excitation Raman difference spectroscopy (SE-SERDS) system for label-free analysis of circulating cell-free DNA (cfDNA). Spectra were acquired from 90 recurrent and 90 non-recurrent NPC patients under non-polarized (NP), left-handed circularly polarized (LHCP), and right-handed circularly polarized (RHCP) excitations, alongside a fused LHCP + RHCP dataset. While the combination of circular polarization and surface-enhanced Raman scattering (SERS) uncovers hidden chiroptical signatures of cfDNA, the instrumentation-based SERDS method effectively eliminates residual fluorescence. In contrast to traditional polynomial-fitting (PF) methods, this hardware-driven approach replaces subjective algorithmic assumptions with physical wavelength modulation, improving apparent spectral resolution and enabling the extraction of subtle prognostic features at 1074 and 1147 cm–1 with a higher signal-to-background contrast. Furthermore, circularly polarized excitation yields more statistically significant spectral differences between groups than NP excitation, with fused LHCP + RHCP spectra providing higher discriminability by capturing complementary conformational information of the cfDNA double helix than standalone polarization data. Using 5-fold cross-validated linear discriminant analysis, SE-SERDS consistently outperformed PF-SERS across all excitation modes, with prediction accuracies rising from 71.1 vs 66.7% (NP) to a maximum of 93.9 vs 87.8% (LHCP + RHCP). Integration with a support vector machine (SVM) further enhanced the accuracy to 95.6%. These results indicate that the fused dual-polarization SE-SERDS strategy provides a powerful tool for pretreatment NPC recurrence prediction to support clinical intervention.