Chemometric Analysis of a Ternary Mixture of Caffeine, Quinic Acid, and Nicotinic Acid by Terahertz Spectroscopy

Abstract

Caffeine, quinic acid, and nicotinic acid are among the significant chemical determinants of coffee quality. This study develops a chemometric model to quantify these compounds in ternary mixtures analyzed by terahertz time-domain spectroscopy (THz-TDS). A data set of 480 THz spectra was obtained from 80 samples. Combinations of data preprocessing methods, including normalization (Z-score, min-max scaling, Mie baseline removal) and dimensionality reduction (principal component analysis (PCA), factor analysis (FA), independent component analysis (ICA), locally linear embedding (LLE), non-negative matrix factorization (NMF), isomap), and prediction models (partial least-squares regression (PLSR), support vector regression (SVR), multilayer perceptron (MLP), convolutional neural network (CNN), gradient boosting) were analyzed for their prediction performance (totaling to 4,711,685 combinations). Results show that the highest quantification performance was achieved at a root-mean-square error of prediction (RMSEP) of 0.0254 (dimensionless mass ratio), using min-max scaling and factor analysis for data preprocessing and multilayer perceptron for prediction. Effects of preprocessing, comparison of prediction models, and linearity of data are discussed.

Figure 1

Chemical structure and content of caffeine, d-(−)-quinic acid, and nicotinic acid in coffee.

Figure 2

Ternary diagram of 80 sample compositions analyzed by THz-TDS. To read the mixture compositions (dimensionless mass ratio), the key on the top right can be placed on a point, extending the lines to the axes or triangle edges. Preparation methods are described in Section 4.1.

Figure 3

THz absorption spectra of pure caffeine, quinic acid, nicotinic acid, and a ternary mixture before (left) and after (right) Mie baseline removal (a.u. = arbitrary unit).

Figure 4

Pipeline of chemometric methods. See the list of abbreviations in Table 4.

Figure 5

Performance of MLP with min-max scaling and FA preprocessing: predicted vs actual (dimensionless) mass ratio of caffeine (red), quinic acid (green), and nicotinic acid (blue) in ternary mixtures (point colors are proportional to their associated predicted composition—see Figure 2).

Figure 6

Prediction performance of nonlinear (blue) and linear (red) MLP models for (dimensionless) mass ratios of caffeine, quinic acid, and nicotinic acid in ternary mixtures. The left vertical axis is shifted up with respect to the right vertical axis for ease of comparison.

Figure 7

Custom-made THz-TDS system at NECTEC, Pathum Thani, Thailand. The photoconductive antenna transmitter and receiver are labeled as Tx and Rx, respectively. Off-axis parabolic mirrors are labeled as #1, #2, #3, and #4.