Ultraviolet spectroscopy and supervised pattern recognition methods for authentication of transgenic and non-transgenic soybean oils

Abstract

A methodology was developed for the authentication of transgenic and non-transgenic soybean oil samples by using Ultraviolet (UV) spectroscopy coupled with Partial Least Squares Discriminant Analysis (PLS-DA). The accuracy, represented by RMSEC and RMSEP, was 0.223 and 0.278, respectively. For the model, the sensitivities were 0.875 and 1.000 for transgenic and non-transgenic classes, respectively. Based on these results, the model was able to classify non-transgenic soybean oil samples. The transgenic class presents specificity equal to 1, this result means that any non-transgenic sample was classified in the transgenic class. The non-transgenic class presents specificity equal to 0.875 due to the prediction of two samples of the transgenic class in the non-transgenic class. The ability of UV spectroscopy for soybean oil authentication can be assigned to the bathochromic shift, probably due to the differences in the chromophore group of genotypic structure present in the transgenic and non-transgenic samples.