Instrument-agnostic multivariate models from normal phase liquid chromatographic fingerprinting. A case study: Authentication of olive oil

The application of non-targeted analytical strategies such as instrumental chromatographic fingerprinting is commonly applied in the field of food authentication/food quality. Although the multivariate methods developed to date are able to solve any authenticity problem, they remain dependent on the instrument state where the signals were acquired, which difficult their transfer to other laboratories. The aim of this research is to develop multivariate models independent of both instrument state and time at which the signals were acquired. For this, chromatograms obtained from the polar fraction of different olive oil samples analysed by (NP)UHPLC-UV/Vis are transformed to instrument-agnostic fingerprints. Instrument independence is achieved by transferring the chromatographic behaviour of an 'ad-hoc' external standards mixture solution analysed throughout an analysis sequence to the remaining analysed samples.& nbsp;The SIMCA models developed from the chromatographic fingerprint matrix before and after instrumentagnostizing showed significant differences in the number of samples classified as inconclusive , with the after model showing the best results. Furthermore, the PLS-DA and SVM models obtained before and after signal instrument-agnostizing showed similar outcomes. The main conclusion of the work has been to verify that the instrument-agnostizing methodology could allow the building of multivariate classification models which could be transferred among different laboratories as they are not influenced by the signal acquisition time.

Automated summary

This research aims to develop multivariate models independent of both instrument state and signal acquisition time for the field of food authentication/food quality. By using instrument-agnostic methods, multivariate classification models can be built and transferred among different laboratories.