LC-HRMS fingerprinting and chemometrics for the characterization and classification of Lotus cultivars from Uruguay: a study on phenolic composition

When defining four classes (<i>L. uliginosus</i>, <i>L. corniculatus</i>, and the two hybrids), the optimal PLS-DA model required six latent variables and achieved 100% classification accuracy, with both sensitivity and specificity reaching 100%. Additional PLS-DA models were developed to assess intra-species discrimination among the 3 <i>L. uliginosus</i> and 5 <i>L. corniculatus</i> cultivars, with varying degrees of separation observed. In each PLS-DA model, VIP loadings scores allowed the selection of the most discriminant phenolic compounds for each class under study. A total of 105 compounds, including phenolic acids, flavonols, flavan-3-ols, proanthocyanidins, and organic acids, were tentatively identified by analyzing all cultivars.