A comparison of several second-order algorithms for simultaneous determination of neomangiferin and mangiferin with severe spectral overlapping in Anemarrhenae Rhizoma

Abstract This work presents a greener approach for simultaneous determination of neomangiferin and mangiferin, the major bioactive constituents with severe spectral overlapping in Anemarrhenae Rhizoma, combining the sensitivity of molecular fluorescence and the selectivity of chemometric multivariate calibration algorithms. In this study, we compared the analytical performance of two group chemometric algorithms including trilinear algorithms such as parallel factor analysis (PARAFAC), alternating trilinear decomposition (ATLD), self-weighted alternating trilinear decomposition (SWATLD) and alternating penalized trilinear decomposition (APTLD), and PLS-based methods such as unfolded partial least-squares or the multi-dimensional partial least-squares, both combined with residual bilinearization (U-PLS/RBL, N-PLS/RBL). The statistical parameters for the validation set of the second calibration were evaluated through the relative error of prediction (REP%), the average recovery (Rec%), and the root mean square error of prediction (RMSEP). Prediction results for the validation set by trilinear algorithms showed that the values were satisfactory for neomangiferin, and higher and not acceptable values for mangiferin, while U-PLS and N-PLS predictions were very successful for two analytes. Therefore, U- and N-PLS/RBL were chosen to determine neomangiferin and mangiferin in more complex real samples simultaneously, and U-PLS/RBL algorithm s howed the best performance. The predicted concentrations by proposed methods were satisfactorily compared with those obtained using high performance liquid chromatography with ultraviolet detection.