Predicting Drugs in Synthetic Oral Fluid Using Handheld Surface-Enhanced Raman Spectroscopy and Multivariate Regression Algorithms

Abstract

Drug detection in oral fluid is important for many applications such as roadside drug testing. Oral fluid in such cases serves as a non-invasive matrix that does not require a specialist for sample collection. Handheld Raman spectroscopy offers a chemically specific and rapid technique for detecting drugs on the spot. However, this technique lacks sensitivity when applied to detecting drugs in biological matrices. Surface-enhanced Raman spectroscopy (SERS) addresses this limitation by enhancing the Raman signal of drugs upon conjugating the drug of interest with metallic nanoparticles. In this work, handheld SERS was combined with multivariate regression for detecting paracetamol, cocaine, and benzoylecgonine in synthetic oral fluid. Solutions were prepared across the studied concentration range down to 0.01 mg/mL, which was the lowest level that yielded identifiable SERS bands. For surface enhanced Raman spectroscopy enhancement, gold and silver substrates were used. Spectra were collected using the Agilent Resolve handheld Raman spectrometer equipped with 830 nm laser wavelength. The spectral data were exported to MATLAB, where principal component regression (PCR) and partial least squares regression (PLSR) were applied. The results showed that optimal SERS enhancement was obtained using a gold substrate for paracetamol and benzoylecgonine and a silver substrate for cocaine. In all cases, both PCR and PLSR models demonstrated high accuracy and precision in predicting the concentration of drugs in synthetic oral fluid, with PLSR outperforming PCR. In summary, handheld SERS predicted the concentrations of the three drugs in synthetic oral fluid non-destructively.