Recognition of samples of similar composition using kinetic-based optical fingerprinting strategy: Choice of...

Optical fingerprinting offers effective solutions to many practical problems that require distinguishing samples of similar composition. We are developing a kinetic version of optical fingerprinting based on conducting an indicator reaction in the presence of a sample and monitoring time-resolved optical changes during the reaction, which generates characteristic sample profiles that can be used for chemometric discrimination. This study investigates the influence of the choice of indicator reaction on the discrimination accuracy of various sample types, including irradiated foods (chicken breasts, potato tubers, and turkey liver), apple juices from different production years, natural waters, and model mixtures of pharmaceuticals. The method involved the oxidation of four heptamethine carbocyanine dyes by NaOCl or H₂O₂ in a 96-well plate with periodic absorption and emission measurements using a photographic method. The resulting kinetic profiles of the samples were processed using partial least squares - discriminant analysis (PLS-DA), linear discriminant analysis (LDA), and logistic/softmax regression. At least one indicator reaction provided high discrimination accuracy (above 90%) for each sample type, but no single reaction was universally optimal for all tasks. PLS-DA proved to be the most effective method for data processing for binary classifications. The kinetic-based methods were compared with alternative approaches, including intrinsic spectra, antioxidant assay (FRAP), and 3,3',5,5'-tetramethylbenzidine (TMB) oxidation, with all methods tested on the same samples.​ The method is versatile and cost-effective, requires simple setup and minimal sample preparation, and is sensitive to minor compositional changes.