Developing New Diagnostic Tools Based on SERS Analysis of Filtered Salivary Samples for Oral Cancer Detection

Abstract

Cancer still represents one of the biggest challenges in current medical practice. Among different types of cancer, oral cancer has a huge impact on patients due to its great visibility, which is more likely to create social stigma and increased anxiety. New early diagnose methods are still needed to improve treatment efficiency and patients' life quality. Raman/SERS (Surface Enhanced Raman Spectroscopy) spectroscopy has a unique and powerful potential for detecting specific molecules that can become priceless biomarkers in different pathologies, such as oral cancer. In this study, a batch of saliva samples obtained from a group of 17 patients with oro-maxillofacial pathologies compared with saliva samples from 18 healthy donors using the aforementioned methods were evaluated. At the same time, opiorphin, potassium thiocyanate and uric acid were evaluated as potential specific biomarkers for oro-maxillofacial pathologies using multivariate analysis. A careful examination of SERS spectra collected on saliva samples showed that the spectra are dominated by the vibrational bands of opiorphin, potassium thiocyanate and uric acid. Given the fact that all these small molecules are found in very small amounts, we filtrated all the samples to get rid of large molecules and to improve our analysis. By using solid plasmonic substrates, we were able to gain information about molecular concentration and geometry of interaction. On the other hand, the multivariate analysis of the salivary spectra contributed to developing a new detection method for oral cancer.

Keywords: RAMAN/SERS spectroscopy; biomarkers; multivariate analysis; opiorphin; oral cancer; plasmonic substrates; potassium thiocyanate; uric acid.

Figure 1

Mean Raman spectra of filtered salivary samples collected from controls (green spectrum) and oral cancer patients (red spectrum), using an excitation wavelength of 785 nm. The main vibrational peaks are marked for clarity.

Figure 2

Mean SERS spectra of filtered salivary samples collected from controls (jade spectrum) and cancer patients (magenta spectrum), using an excitation wavelength of 785 nm. The main vibrational peaks are marked for clarity.

Figure 3

Classification performance obtained after the PCA-LDA analysis of the full range SERS spectra using 5 PC components. AUROC—the area under the receiver operating characteristic. 77% accuracy, 71% sensitivity and 83% specificity were obtained.

Figure 4

Classification performance obtained after the PCA-LDA analysis of the SERS spectra in the 550–1250 cm⁻¹ spectral interval using 5 PC components: 74% accuracy, 53% sensitivity and 94% specificity were obtained.

Figure 5

Classification performance obtained after the PCA-LDA analysis of the SERS spectra in the 1950–2150 cm⁻¹ spectral interval using 5 PC components: 74% accuracy, 71% sensitivity and 78% specificity were obtained.

Figure 6

Classification performance obtained after a PCA-LDA analysis of the full range Raman spectra using 5 PC components: 57% accuracy, 53% sensitivity and 61% specificity were obtained.