Front-face fluorescence spectroscopy combined with chemometrics for the discrimination of wheat flour and cassava flour

Abstract

Fluorescence and chemometric methods were used in this study to discriminate wheat flour samples mixed with cassava flour at different percentages (from 0 to 60). The fluorescence emission spectra of the different flours were obtained at excitation wavelengths of 349 nm and 373 nm. Principal component analysis (PCA) and Cluster Analysis (CA) showed that pure wheat flour can be distinguished from mixed wheat flour. The overall predictive accuracy of the principal component analysis and discriminant analysis (PCA-DA) and partial least squares discriminant analysis (PLS-DA) of the two-class discriminant models (pure and mixed wheat) reached 100%, showing that the model was able to correctly differentiate a pure wheat sample from a mixed wheat sample. The predictions of the three-class models (pure wheat, 10%,20%−60%; pure wheat, 10%+20%, 30%−60% and pure wheat, 10%−30%, 40%−60%) showed an overall accuracy of over 90% for the PCA-DA and PLS-DA models, the latter being the best performing. Thus, the specificity, sensitivity and precision values (>0.8) for the 10%, 10% + 20% and 10%− 30% classes in each PLS-DA model showed the ability of these models to predict the classes of blended wheat flour samples according to their level of mixing by cassava.