Website: https://vjfc.nifc.gov.vn/
Website: https://vjfc.nifc.gov.vn/
This study has developed a rapid and simple method based on the UV-Vis full spectra coupled with machine learning for the quantitative prediction of curcumin, demethoxycurcumin and bisdemethoxycurcumin in turmeric rhizomes. The UV-Vis spectral data and HPLC quantification results of the individual components from 55 turmeric rhizome samples were utilized for model development and training. Analogous data matrices from an independent set of 24 samples were used for external validation of the developed models. Four machine learning models were investigated, comprising two linear multivariate regression algorithms: principal component regression (PCR) and partial least squares regression (PLSR), two non-linear multivariate regression algorithms: artificial neural network (ANN) and random forest (RF). The results demonstrated that the linear multivariate regression models exhibited superior analytical performance. Specifically, PCR yielded a coefficient of determination (R²) values from 0.957 to 0.982 with root mean square error (RMSE) values from 3.086 to 1.295, while PLSR achieved R² values from 0.956 to 0.979 and RMSE values from 3.116 to 1.139. However, when comparing the HPLC-quantified contents with the values predicted by the two models, some samples still exhibited relative errors exceeding 20%. This study confirms the feasibility of rapidly and simultaneously predicting the contents of curcumin, demethoxycurcumin and bisdemethoxycurcumin in turmeric rhizomes using UV-Vis spectral data coupled with either PLSR or PCR models, offering an efficient alternative to conventional methods.
Machine Learning, UV-Vis, curcumin, turmeric.
[1]. D.H Bich, D.Q. Chung, B.X. Chuong, "The medicinal plants and animals in Vietnam," Hanoi for science and technology, vol. 2, pp. 1256, 2006 [in Vietnamese].
[2]. A. M. Araya-Sibaja, F. Vargas-Huertas, S. Quesada et al., “Characterization, antioxidant and cytotoxic evaluation of demethoxycurcumin and bisdemethoxycurcumin from Curcuma longa cultivated in Costa Rica,” Separations, vol. 11, no. 1, pp. 23, 2024.
[3]. V. S. R. Kotra, L. Satyabanta, and T. K. Goswami, “A critical review of analytical methods for determination of curcuminoids in turmeric,” Journal of Food Science and Technology, vol. 56, no. 12, pp. 5153-5166, 2019.
[4]. T. T. Hue, T. T. T. Dung, N. V. Ri et al., “Simultaneous determination of caffeine, theobromine, theophylline in tea using ultraviolet-visible spectroscopy combined with multivariate analysis,” Journal of Analytical Sciences, vol. 24, no. 1, pp. 192-196, 2019 [in Vietnamese].
[5]. A. Alqahtani, T. Alqahtani, A. A. Fatease et al., “Rapid UV-Vis spectrophotometric method aided by firefly-PLS models for the simultaneous quantification of ciprofloxacin, lomefloxacin, and enrofloxacin in their laboratory mixture, dosage forms and water samples: greenness and blueness assessment,” BMC Chemistry, vol. 18, no. 1, 2024.
[6]. R. Khani, R. Rahmanian, and N. V. Motlagh, “UV-Visible spectrometry and multivariate calibration as a rapid and reliable tool for simultaneous quantification of ternary mixture of phenolic acids in fruit juice samples,” Food Analytical Methods, vol. 9, no. 5, pp. 1112-1119, 2016.
[7]. R. Ergon, “Principal component regression (PCR) and partial least squares regression (PLSR),” in Mathematical and Statistical Methods in Food Science and Technology, pp. 143-174, 2013.
[8]. K. Upreti, M. Verma, M. Agrawal et al., “Prediction of mechanical strength by using an artificial neural network and random forest algorithm,” Journal of Nanomaterials, vol. 2022, no. 1, 2022.
[9]. D. Chicco, M. J. Warrens, and G. Jurman, “The coefficient of determination R-squared is more informative than SMAPE, MAE, MAPE, MSE and RMSE in regression analysis evaluation,” PeerJ Computer Science, vol. 7, 2021.