Figure 3

PCA and PLS-DA analysis; principal component analysis PCA score plot (a) showing distinct relationship between samples and differences in KATR versus KANTR samples along x-axis (PC1) and within groups along y-axis (PC2). PC1 and PC2 showed variance at 33.2% and 22.1% respectively at 95% confidence interval; partial least squares-discriminant analysis (PLS-DA) Scores plot of metabolite features (b) in PLS-DA model representing covariance between component 1 (32%) at x axis and component 2 (17.5%) at y axis; (c) cross validation of PLD-DA model with positive Q2 reflecting predictability and non-overftting of the model and top 25 annotated metabolites identified from metabolite features with variable importance at x axis. A higher VIP score (> 1.0) for metabolites represent higher importance in influencing the scores most.