partial least-squares discrimination analysis (pls-da) classification method  (MathWorks Inc)

Journal: Metabolomics

Article Title: NMR-based urinary biomarkers in pediatric primary mitochondrial disorders and chronic kidney disease: shared mitochondrial dysfunction, diverging biosignatures

doi: 10.1007/s11306-025-02363-8

Figure Lengend Snippet: Multivariate analysis of urinary ¹H NMR metabolomic profiles across groups: Primary Mitochondrial Disease (PMD, orange), Suspected Mitochondrial Disease (SMD, yellow), Control individuals (green), Chronic Kidney Disease (CKD) stages 1–2 (lilac) and stages 3–5 (purple). The ellipses represent 95% confidence intervals. A Principal Component Analysis (PCA) highlights distinct metabolic fingerprint in CKD patients compared to other groups. B Corresponding supervised model using partial least squares discriminant analysis (PLS-DA) shows metabolic divergence between groups, particularly between PMD, CKD3-5, and controls. PMD primary mitochondrial disease, SMD suspected mitochondrial dysfunction, CKD chronic kidney disease

Article Snippet: Multivariate analysis (MVA) trough Principal component analysis (PCA) and partial-least-squares discriminant analysis (PLS-DA) models (SIMCA-P 11.5, Umetrics, Umeå, Sweden), were performed on full resolution 1 H NMR spectra, after unit variance (UV) scaling of the spectra.

Techniques: Metabolomic, Control

Journal: Metabolomics

Article Title: NMR-based urinary biomarkers in pediatric primary mitochondrial disorders and chronic kidney disease: shared mitochondrial dysfunction, diverging biosignatures

doi: 10.1007/s11306-025-02363-8

Figure Lengend Snippet: Supervised multivariate models and spectral features associated with mitochondrial disorders. I. Partial Least Squares Discriminant Analysis (PLS-DA) models were generated to compare patients with PMD to: A controls, B SMD, C . CKD stages 1–2, and D . CKD stages 3–5. The corresponding Q² values, indicate the predictive performance of each model. Clear discrimination was observed between PMD and controls (Q² = 0.53), and both CKD subgroups (Q² = 0.62 and 0.78). In contrast, the model comparing PMD to SMD showed lower predictive performance (Q² = 0.25), reflecting significant metabolic overlap. II. Corresponding loadings plot of model A, color-coded by variable importance in projection (VIP) scores for the first latent variable (LV1) highlighting spectral regions that contribute most to the separation between PMD and controls. (regions around 1.2–2.5 ppm and 3.0–4.5 ppm, consistent with metabolites involved in amino acid metabolism, organic acids, and energy-related pathways). Three-letter code used for amino acids. Abbreviations: 3-HBA: 3-hydroxybutyrate; 4-HPA: 4-hydroxyphenylacetate; N -PhAcGly: N -phenylacetylglycine. Ui: resonances at chemical shift i, which remain unassigned (or with tentative assignment)

Article Snippet: Multivariate analysis (MVA) trough Principal component analysis (PCA) and partial-least-squares discriminant analysis (PLS-DA) models (SIMCA-P 11.5, Umetrics, Umeå, Sweden), were performed on full resolution 1 H NMR spectra, after unit variance (UV) scaling of the spectra.

Techniques: Generated