Standardized analytical multivariate methods via agnostization of LF-NMR signals - Updating extra virgin olive oil...

Extra virgin olive oil (EVOO) is a high-value commodity traded worldwide; thus, accurate quality control analyses are essential to ensure its genuineness. In this context, a universal analytical methodology capable of delivering precise, fast and robust quality assessments would be highly beneficial for stakeholders. This study extends the instrument-agnostizing methodology from chromatography to low-frequency field nuclear magnetic resonance spectroscopy (LF-NMR). The proposed approach outlines the required steps to generate instrument-agnostic fingerprints and demonstrates its application through the development of a global multivariate model. For this purpose, 93 edible vegetable oils (76 EVOO and 17 non-olive ones) and 9 vegetable oil blends were analyzed using two different LF-NMR spectrometers operating at 100 MHz and 80 MHz ¹H field frequencies. Each LF-NMR signal was subjected to the instrument-agnostizing methodology to obtain harmonized LF-NMR fingerprints. To demonstrate the usefulness of this methodology, principal component analysis was performed to assess the instrument-related variability before and after the instrument-agnostizing methodology. Finally, partial least squares-discriminant analysis models were developed for each dataset of LF-NMR fingerprints and evaluated using four main classification performance metrics (sensitivity, specificity, precision and accuracy). The developed classification model with 100 MHz LF-NMR fingerprints achieved the required values of quality performance metrics in a conformity scenario, showing for external and cross-instrument validation sets a sensitivity of 96 % in both cases, and highlighting a precision of 100 % and 96 %, respectively. The excellent outcome highlights the viability of the instrument-agnostizing methodology, enabling consistent and accurate quality control across laboratories, regardless of the LF-NMR instrument used.