Incorporation of Partial Least Squares-Discriminant Analysis with Ultra-High-Performance Liquid Chromatography Diode-Array Detector for Authentication of Skin Gelatine Sources

Abstract

This research seeks to (1) authenticate sources of skin gelatine by combining putative 17 amino
acids (AAs) analysis with chemometrics by Ultra-High-Performance Liquid Chromatography
Diode-Array Detector (UHPLC-DAD) and (2) create AA profiles in skin gelatines. The
classification capability of partial least square-discriminant analysis (PLS-DA) models was
assessed to determine the most effective discriminant model. Principal component analysis
(PCA) with quartimax rotation was utilised to accurately organise gelatine clusters and assign the
significantly contributing AAs to each cluster. The PLS-DA model with 13 AAs (PLS-DAVIPAA)
outperformed the PLS-DA model with 17 AAs (PLS-DAAA) because its R2Y (0.938), R2X (0.881),
and Q2 (0.929) values were greater. With 13 significant AAs, the PLS-DAVIPAA model obtained
cluster classification accuracy of 100% on training and cross-validation datasets and 93.3% on
testing and verification datasets. The chemical structure of gelatines may shed light on the
interactions between AAs. Following six quartimax rotations, the gelatines were grouped
correctly. The PCA showed the dominant presence of these AAs: L-Valine, L-Phenylalanine and
L-Tyrosine in porcine gelatine; Glycine, L-Threonine, L-Arginine, L-Methionine, L-Histidine and
L-Serine in fish gelatine; and L-Hydroxyproline, L-Leucine and L-Proline in bovine gelatine. The
authority could use this technique to set a standard for authenticating skin gelatine samples.