Rapid, sensitive detection of ganciclovir, penciclovir and valacyclovir-hydrochloride by artificial neural network and partial least squares combined with surface enhanced Raman spectroscopy

Surface enhanced Raman spectroscopy (SERS) plays important role in qualitative analysis, identification and bio-imaging analysis, but the difficulty of quantitative analysis still exists. Here, we established quantitative prediction models for ganciclovir (GCV), penciclovir (PCV) and valacyclovir-hydrochloride (VACV-HCl) by adopting chemometric methods including artificial neural network (ANN) and partial least squares (PLS) algorithms combined with SERS based on concentrated Ag nanoparticles. The limit of detection for three drugs reached 1.0 x 10(-6) mol L-1 and the detection time was less than 4 min for a single sample, which demonstrated that SERS detection is rapid and sensitive. Comparing with the PLS models, the ANN models established in this paper showed better performance, the root mean square error of prediction and correlation coefficients of prediction for GCV, PCV and VACV-HCl were 0.0009245, 0.0002237, 0.0003307 and 0.8991, 0.9867, 0.9880, respectively. These results indicated that the established ANN models are robust and accurate. Subsequently, the ANN models combined with SERS were applied to detect VACV-HCl tablets and rat plasma spiking GCV and PCV. Overall, chemometrics combined with SERS in this paper provides a new reference for analytes to develop a rapid and sensitive quantitative analysis method.