Interval wavelength selection and simultaneous quantification of spectrally overlapping food colorants by multivariate calibration

Variables selection is often necessary to remove redundant data and to reduce the negative influence of spectral overlapping. In the current work, two interval variable selection methods were applied to quantify five coloring agents (Tartrazine TAR, Sunset Yellow SY, Allura Red AR, Brilliant Blue BBL, and Brilliant Black BBK) which exhibited intense spectral overlapping in powdered soft drinks. Interval partial least squares iPLS and net analyte signal NAS methodology were used to pick up the most informative variables for dyes quantification in powdered soft drinks. Based on NAS calculations, the optimum sensitivity and selectivity for dyes measurement were found to be (1.23–5.56) and (0.30–0.72), respectively and at pH 3.0. Moreover, the minimum spectral overlapping (28–70%) among dyes was observed at pH 3.0 while the maximum overlapping (38–74%) was at pH 10.0. Interval partial least squares iPLS was more capable to handle the overlapping between SY and AR dyes. On the other hand, net analyte signal method was effective to capturing the informative regions for dyes of lower spectral overlapping, TAR, BBL, and BBK. In the case of AR, the best prediction (REP% 2.4) was achieved at 350–530 nm using iPLS. However, the best prediction of BBL (REP% 3.4%) was achieved at 655–680 nm (i.e., 6 variables) using NAS/PLS regression. The ability NAS/PLS regression, which uses fewer number of variables, was attributed to NAS mechanism which extracts the net signal of the analyte, thus; preventing overlapping with the rest of compounds signals and this will elegantly leads to fewer number of variables. The superiority of iPLS to calibrate intensely overlapping dyes is attributed to its inherent mechanism of selecting the spectral data that include all possible variables leading to better prediction. At the optimum calibration conditions, the dyes were detected in powdered soft drinks with adequate accuracy (%recoveries 97.3–107.5) and precision (RSD 3.1–9.30). The maximum total concentration of dyes was reported in orange drink samples reaching to 567 mg/kg. The result highlights and emphasizes the highly required further monitoring of this type of food, considering the damages of such popular synthetic dyes to human health. Analysis of results by ANOVA indicated that the total content of dyes was statistically comparable in the samples while the total content of each single dye was statistically different in tested samples.