High-Accuracy Renal Cell Carcinoma Discrimination through Label-Free SERS of Blood Serum and Multivariate Analysis

Abstract

Renal cell carcinoma (RCC) represents the sixth most frequently diagnosed cancer in men and is asymptomatic, being detected mostly incidentally. The apparition of symptoms correlates with advanced disease, aggressive histology, and poor outcomes. The development of the Surface-Enhanced Raman Scattering (SERS) technique opened the way for investigating and detecting small molecules, especially in biological liquids such as serum or blood plasma, urine, saliva, and tears, and was proposed as a simple technique for the diagnosis of various diseases, including cancer. In this study, we investigated the use of serum label-free SERS combined with two multivariate analysis tests: Principal Component Analysis combined with Linear Discriminate Analysis (PCA-LDA) and Supported Vector Machine (SVM) for the discrimination of 50 RCC cancer patients from 45 apparently healthy donors. In the case of LDA-PCA, we obtained a discrimination accuracy of 100% using 12 principal components and a quadratic discrimination function. The accuracy of discrimination between RCC stages was 88%. In the case of the SVM approach, we obtained a training accuracy of 100%, a validation accuracy of 92% for the discrimination between RCC and controls, and an accuracy of 81% for the discrimination between stages. We also performed standard statistical tests aimed at improving the assignment of the SERS vibration bands, which, according to our data, are mainly due to purinic metabolites (uric acid and hypoxanthine). Moreover, our results using these assignments and Student’s t-test suggest that the main differences in the SERS spectra of RCC patients are due to an increase in the uric acid concentration (a conclusion in agreement with recent literature), while the hypoxanthine concentration is not statistically different between the two groups. Our results demonstrate that label-free SERS combined with chemometrics holds great promise for non-invasive and early detection of RCC. However, more studies are needed to validate this approach, especially when combined with other urological diseases.

Label-Free SERS of Urine Components: A Powerful Tool for Discriminating Renal Cell Carcinoma through Multivariate Analysis and Machine Learning Techniques

“…Another small molecule with a specific SERS signature usually detected in biological fluid is uric acid. As demonstrated in our previous works [32], there are two intense peaks associated with uric acid in the case of blood serum recorded on our solid plasmonic substrate at 640 cm −1 and 1136 cm −1 , and other medium intensity peaks recorded at 496 cm −1 , 590 cm −1 , 813 cm −1 , and 889 cm −1 . On the other hand, the uric acid contains three carbonyl (C=O) groups, usually characterized by a stretching vibration above 1700 cm −1 readily detected by Raman spectroscopy.…”

“…We also notice a high coefficient of determination among these two latter peaks (0.94). At this point, it is worth mentioning that in the case of serum, the SERS intensity at 640 cm −1 and 1135 cm −1 , assigned to uric acid, correlates with a coefficient of determination higher than 0.9 with the SERS intensities at 812 cm −1 and 889 cm −1 [32], which are situated very close to the 804 cm −1 and 893 cm −1 urine peaks recorded in the present study. We therefore cannot exclude a contribution of the uric acid for these peaks, the 640 cm −1 being probably masked by the intense creatinine vibration at 670 cm −1 .…”

“…Other assignments of the SERS peak of urine recorded around 1000 cm −1 are albumin [42,49,53,54] and phenylalanine [33,41], an amino acid associated with albumin, or its aromatic ring stretching [40]. However, as pointed out earlier, our solid plasmonic platform specifically enhances the Raman signal of small molecules, even in a high protein concentration environment [32]. Moreover, the studies assigning this peak to albumin or its constituents were performed on colloidal SERS platforms sometimes specifically functionalized for the detection of proteins [53].…”

“…In the label-free SERS approach, the spectra of biological liquids strongly depend mainly on molecules affinity for the plasmonic substrate and, to a lesser extent, on their relative concentration. While in the case of colloidal silver and gold nanoparticles the nanoparticles will be surrounded by proteins that will dominate the SERS, in the case of solid plasmonic substrates with a near-infrared excitation wavelength of 785 nm, the most prominent bands observed in SERS mainly correspond to lower molecular mass molecules, even in the case of highly concentrated biofluids, as we recently demonstrated in the case of blood serum [32]. The fact that the solid substrates amplify mostly the Raman signal of small molecules was first observed in a pioneering paper by Premasiri et al [30].…”

“…In previous reports, it was demonstrated that reproducible SERS spectra, obtained through a solid plasmonic SERS substrate, allowed for a possibility of discrimination of breast cancer patients through labelfree SERS of blood plasma, combined with multivariate analysis [31]. Also, we recently proved that label-free SERS of blood serum analyzed either through Principal Component Analysis combined with Linear Discriminate Analysis (PCA-LDA) or through Support Vector Machine (SVM) could discriminate with high accuracy RCC patients from apparently healthy controls [32].…”