Prediction of tyrosinase inhibitory activities of Morus alba root bark extracts from HPLC fingerprints

Abstract

A correlation model between tyrosinase inhibitory activities and HPLC chromatograms of Morus alba root bark extracts was calibrated and validated. The correlational approach has been suggested as an alternative technique for quality control and study of natural products. However, this method requires a large number of various samples with similar chemical compositions. In the present study, 42 different extracts were obtained from 6 samples of M. alba root bark under 7 different extraction solvent conditions using an automated pressurized liquid extraction system. HPLC chromatograms and mushroom tyrosinase inhibitory activities of these extract samples were obtained, and the chromatograms were preprocessed with asymmetric least square smoothing and correlation optimized warping algorithms. The partial least squares (PLS) regression model between the preprocessed chromatograms and the bioactivities of samples was created. The full cross validations were performed by changing the selection method of the test sets; plant sample based, extraction method based validation set selections were used for the validation tests. In both validation tests, the PLS models showed excellent performance of prediction. The predictive models showed averagely RMSEP of 9.4309 and R2 values of 0.8717 in the validation test based on plant samples, and RMSEP of 8.1541 and R2 values of 0.8778 in the validation test based on extraction methods. Additionally, an estimation of the bioactive component was made from the regression coefficient, and the result of the estimation was similar to the experimental results of reference papers.