Study on quality differences among commercial specifications of Coicis seed based on multi-dimensional feature analysis

To reveal quality discrepancies among five commercial Coix seed specifications, we systematically compared their morphology, nutrients and bioactive constituents. Morphological and nutritional indices were quantified; Gas chromatography-mass spectrometry (GC-MS) profiled amino, fatty and organic acids, whereas high performance liquid chromatography- evaporative light scattering detector (HPLC-ELSD) determined triacylglycerols. Orthogonal partial least squares discriminant analysis (OPLS-DA) screened 13 candidate discriminatory markers (VIP > 1), combined with DPPH and ABTS⁺ antioxidant data for multivariate discrimination. An artificial neural network (ANN) achieved 100% training and 86.7% prediction accuracy for sample classification. Obvious inter-specification differences in appearance, chemical profiles and antioxidant capacity were validated. Integrating chemometrics and machine learning, this work identifies characteristic markers to distinguish Coix seed varieties, offering reliable theoretical support for germplasm assessment, quality classification and functional exploitation of Coix seeds.