The multivariate calibration methods—principal component regression (PCR) and partial least squares (PLSs)—were employed for the prediction of total phenol contents of four
The antioxidant activity could be dependent on the extraction solvent, the hydrophilicity of compounds, the sample, and type of phenolic compounds, which means that different phenolic compounds react in different ways in antioxidant activity assays. The phenolic compounds prove the importance of antioxidant behaviour and contribute significantly to the total antioxidant activity of medicinal and aromatic samples [
Multivariate chemometric methods such as principal component regression (PCR) and partial least squares (PLSs) allow to extract analytical information from the full spectra/chromatograms, providing to use simultaneously an elevated number of signals. Moreover, these techniques allow a rapid analytical response with minimum sample preparation, reasonable accuracy and precision, and without a preliminary separation step in complex matrices [
Preprocessing of chromatographic data is necessary for complex mixtures and chemometrics offers many tools well suited to handling this task. In addition, it should be emphasized that a powerful application of multivariate calibration methods requires a careful preprocessing of the chromatograms. Therefore, different preprocess techniques such as baseline correction to enhance the signal to noise ratio [
In the present study, PCR and PLS multivariate calibration models were developed for the prediction of total phenol content of
Rosmarinic acid was purchased from Sigma-Aldrich (St. Louis, USA), the Folin-Ciocalteu phenol reagent, gallic acid, quercetin, kaempferol, and rutin were purchased from Sigma and were used without further purifications. Analytical grade of hydrochloric acid, HPLC grade of methanol, butanol, ethyl acetate, acetonitrile, hexane, formic acid, and caffeic acid were purchased from Merck (Merck, Darmstadt, Germany).
The collected samples were dried at room temperature and stored at 4°C. The whole parts of
The HPLC analysis of
The total phenolic content by the Folin-Ciocalteu reagent was determined according to the procedure reported in the literature [
The data set containing 96 chromatograms of 48
To select the optimum number of factors to be used in the PLS and PCR calibration models a cross-validation procedure was used. In this procedure, each
The performance of the calibration model and its prediction ability is measured by the root mean square error (RMSE) obtained on the calibration set and root mean square error of prediction (RMSEP) obtained on the test set, respectively,
Chromatographic profiles can be organized in an
Chemometric treatment of the chromatogram requires that all signals are adjusted to the same length and that corresponding variables (such as peak apexes) are placed in proper columns of the data matrix. Correlation optimized warping aligns chromatograms by piecewise linear stretching and compression of the time axis. At the beginning of the procedure, the profile to be aligned (
Data smoothing techniques are used to eliminate “noise” and extract real trends and patterns. One of the smoothing techniques, Whittaker smoothing was applied for the noisy data (chromatogram) in this study [
Principal components are primarily abstract mathematical entities and further details are described in the literature [
Partial least squares is often presented as the major regression technique for multivariate data to express the relation between
Total phenol contents were determined in
Total phenol contents of
Extraction solvent | ||||
---|---|---|---|---|
Water | ||||
Methanol | ||||
Butanol | ||||
Acetonitrile | ||||
Ethyl acetate | ||||
Hexane | ||||
Acidic water | ||||
Acidic methanol | ||||
Acidic butanol | ||||
Acidic acetonitrile | ||||
Acidic ethyl acetate | ||||
Acidic hexane |
Mean of two determinations ± SD (standard deviation).
Multivariate calibration models were built with data matrix
RMSEs (mg GAE g−1 dried plant) for the calibration and test sets for the total phenol contents of
Model | Preprocessing | Number of components | RMSE | RMSEP |
---|---|---|---|---|
PCR | Column centering | 6 | 6.54 | 6.62 |
Smoothing, normalization, and column centering | 6 | 6.63 | 4.11 | |
PLS | Raw data | 6 | 2.47 | 1.61 |
Smoothing and normalization | 6 | 4.83 | 2.76 |
In the case of shifts present in retention times between chromatograms, alignment of the corresponding peaks is needed. Therefore, the correlation optimized warping was performed to align the chromatograms. It was found that column centering gave the best results for the calibration set, which has the lowest RMSEP. All other results were obtained by data preprocessed in this way. The RMSE of the column centered data is similar to the RMSE of data preprocessed with smoothing, normalization and column centering for PCR calibration. However, the RMSE of column centered data is smaller than the RMSE of data preprocessed with smoothing, normalization and column centering for PLS calibration. Preprocessing with column centering, smoothing followed by normalization and column centering were successfully applied for the warped HPLC chromatograms. It is seen that the PLS model allows better prediction than PCR model for column centered data (Figure
Total phenol contents of (a) the principal component regression model (six factors) (b) the partial least squares regression model (six factors) built with 36 chromatograms.
PLS components were evaluated to verify whether the extracts with high total phenol content could be distinguished on the score plot. When examining the score plot (Figure
Score plot of samples obtained by PLS calibration.
One of the major aspects of this study is to identify the compounds in the plant extracts potentially responsible for the total phenol content of the samples. The PLS loadings were evaluated to investigate the contribution of the phenolic compounds to the total phenol content (Figure
Loadings plot of variables obtained by PLS calibration.
HPLC chromatogram of acidic water extract of
Individual contribution of phenolic compounds to the total phenol content was investigated in order to explain positive and negative PLS loadings. Levels of total phenol content of individual phenolic compounds and mixtures at 1.66 mM were ranged between 0.267
Total phenol content of standard phenolic compounds that were added in the acidic methanol extract of
Phenolic compounds | Total phenol content |
---|---|
Extract | |
Rosmarinic acid | |
Rutin | |
Quercetin | |
Caffeic acid | |
Kaempferol | |
Mixture | |
Mixture (without rosmarinic acid) | |
Mixture (without rutin) | |
Mixture (without quercetin) | |
Mixture (without caffeic acid) | |
Mixture (without kaempferol) |
GAE: gallic acid equivalent.
Mean of two determinations ± SD (standard deviation).
According to the overview of the results, PCR and PLS calibration models were constructed to model the total phenol content of the
The authors are thankful to Uludag University Research Foundation (Project no. 2009/38) for providing financial support for this study.