Partial least squares methods in open-source software for identification of adulterated creatine using infrared...

Creatine is one of the most consumed food supplements by people who practice physical activity. We developed a new methodology by associating Attenuated Total Reflection-Fourier Transform Infrared (ATR-FTIR) with multivariate analyses for screening adulterate creatine. We built partial least squares regression (PLS) and partial least squares discriminant analysis (PLS-DA) models to detect creatine adulteration with corn starch (CS). All models were developed using the open-source software GNU Octave. Asymmetric stretching bands related to carboxyl group present in the creatine molecule and stretching of glycosidic bonds of CS were essential for distinguishing adulterated creatine. PLS showed a strong experimental fit with RMSEp of 9.15 %. PLS-DA proved effective in distinguishing pure and adulterated creatine samples with an accuracy of 97 %. In blind prediction applied to other samples of local trade, the model achieved 95 % accuracy, 90 % sensitivity and 100 % specificity. These results confirm the ability of the employed models to discriminate adulterated creatine.