Figure 2.

Two-dimensional partial least squares discriminant (2D-PLS-DA) scores plots generated from polar metabolite profiles including (A) all capture methods including dart (orange), dropnet (blue), and helicopter (helo, green) methods; (B) dart versus dropnet; (C) dart versus helicopeter capture methods; and (D) helicopter and dropnet captures methods. The number of animals in each group consisted of n = 73 for dart; n = 93 for dropnet; and n = 375 for helicopter capture, with shaded ellipses representing 95% confidence intervals. PLS-DA validation metrics included (i) Q² = 0.80 (component 5), R² = 0.82 (component 5), classification error rate (CER) < 0.10 (component 5), Area under ROC Curve (AUC) > 0.95 (Dart vs. others and Helo vs. others) (component 1), (ii) Q² = 0.85 (component 5), R² = 0.87 (component 5), CER < 0.03 (component 5), AUC = 0.95 (component 1), (iii) Q² = 0.86 (component 5), R² = 0.89 (component 5), CER < 0.02 (component 5), AUC = 0.99 (component 1), and (iv) Q² = 0.62 (component 5), R² = 0.69 (component 5), CER < 0.04 (component 4), AUC > 0.96 (component 1). Permutation tests (n = 2000) for all PLS-DA models p < 0.001.