Near Infrared Spectroscopic Combined with Partial Least Squares and Radial Basis Function Neural Network to Analyze Paclitaxel Concentration in Rat Plasma.

Abstract

Paclitaxel is known as one of the most effective anticancer drugs. Near Infrared Spectroscopy (NIRS), a rapid, precise and non-destructive approach of analysis, has been widely used for qualitative and quantitative detection. The present study aims to analyze the plasma paclitaxel concentration with NIRS. Various batches of plasma samples were prepared and the concentration of paclitaxel was determined via high performance liquid chromatography tandem mass spectrometry (LC-MS/MS). The outliers and the number of calibration set were confirmed by Monte Carlo algorithm combined with partial least squares (MCPLS). Since NIR spectra may be contaminated by signals from background and noise, a series of preprocessing were performed to improve signal resolution. Moving window PLS and radical basis function neural network (RBFNN) methods were applied to establish calibration model. Although both PLS and RBFNN models are well-fitting, RBFNN-established model displayed better qualities on stability and predictive ability. The correlation coefficients of calibration curve and prediction set (Rc(2) and Rp(2)) are 0.9482 and 0.9544, respectively. Moreover, independent verification test with 20 samples confirmed the well predictive ability of RBFNN model.