An Integrated Authentication Analysis of Citrus aurantium L. Essential Oil Based on FTIR Spectroscopy and Chemometrics with Tuning Parameters

Abstract

Citrus aurantium L. essential oil or orange oil (OO) became more popular in recent years due to its benefit for human health. An “economically motivated adulteration” can be potentially occurred to achieve more profit in the market. On the other hand, a cheaper oil such as coconut oil (CO) was commonly used as adulterant. The objective of this study was to perform authentication analysis of OO by FTIR spectroscopy and chemometrics. Principal component analysis was applied in the exploratory data analysis at the initial stage of authentication analysis. Multivariate calibration of principal component regression (PCR) and partial least squares regression (PLSR) were constructed using five types of pre-processed FTIR spectral data. The PCR model using Standard Normal Variate (SNV) spectra was selected as the best prediction model for OO (Rcal2 = 0.999; RMSEC = 0.193; RCV2 = 0.998; RMSECV = 0.456; Rval2 = 0.992; RMSEP = 0.989), whereas the PLSR model using SNV spectra was selected as the best prediction model for CO (Rcal2 = 0.999; RMSEC = 0.174; RCV2 = 0.999; RMSECV = 0.476; Rval2 = 0.992; RMSEP = 0.991). SNV spectra of OO, CO, and binary mixture of OO+CO were used to generate sparse partial least squares-discriminant analysis (sPLS-DA) model. Tuning parameters of component numbers, the number of variables “keepX”, and the distance of prediction was executed. The component number of three with “keepX” for component 1, 2, and 3 were 1, 5, and 1, respectively, were selected along with the maximum distance approach to construct the discriminant model. The final sPLS-DA model explained the total variances of 94% with satisfaction separatibility of 100%, 97.8%, and 100% for OO, CO, and OO+CO, respectively.