Quantitative prediction of a functional ingredient in apple using Raman spectroscopy and multivariate calibration analysis

We propose a method for predicting the concentration of a functional ingredient, procyanidin, in apple using Raman spectroscopy in combination with multivariate calibration analysis. A regression model was constructed by partial least-squares (PLS) regression using the collected Raman spectra and the procyanidin concentrations measured by high-performance liquid chromatography (HPLC). Four different preprocessing algorithms—baseline correction, noise removal, averaging, and multiplicative scatter correction—were applied to the acquired Raman spectra. HPLC was used to determine the procyanidin concentrations in the edible part of apple specimens. The PLS regression model predicted the procyanidin concentration in apple with a coefficient of determination of 0.74, a root-mean-square error of calibration of 7.09 µg/g, and a root-mean-square error of prediction of 14.89 µg/g. In addition, the spectra of the carotenoid pigments were observed from the factors extracted from the PLS analysis. Consequently, we found that the procyanidin concentration in apple can be predicted using Raman spectroscopy measurements of carotenoid pigments of apple peel. Compared with conventional destructive measurements, Raman spectroscopy with the aid of multivariate analysis shows strong potential for the rapid and nondestructive quantitative analysis of procyanidin in apples.