Table 2.
Evaluation of the discrimination among NAG−, CAG+, PLGC and GC metabolic profiles by PLS-DA using cross validated accuracy (i.e. % correctly classified samples), AUROC, sensitivity, specificity, PPV and NPV estimates.
| Model | Latent variables | Accuracy (p-value)1 | AUROC (p-value) | Sensitivity (p-value) | Specificity (p-value) | PPV (p-value) | NPV (p-value) |
|---|---|---|---|---|---|---|---|
| NAG- vs CAG+ | 2 | 0.41 (>0.05) | 0.43 (>0.05) | 0.40 (>0.05) | 0.42 (>0.05) | 0.42 (>0.05) | 0.40 (>0.05) |
| NAG- vs PLGC | 1 | 0.56 (>0.05) | 0.82 (>0.05) | 0.38 (>0.05) | 0.56 (>0.05) | 0.60 (>0.05) | 0.60 (>0.05) |
| CAG + vs PLGC | 1 | 0.53 (>0.05) | 0.49 (>0.05) | 0.57 (>0.05) | 0.50 (>0.05) | 0.54 (>0.05) | 0.53 (>0.05) |
| NAG− vs GC | 2 | 0.77 (0.002) | 0.83 (0.002) | 0.80 (0.002) | 0.74 (0.004) | 0.76 (0.004) | 0.78 (0.004) |
| CAG + vs GC | 2 | 0.80 (0.002) | 0.81 (0.002) | 0.75 (0.008) | 0.85 (0.002) | 0.83 (0.002) | 0.77 (0.006) |
| PLGC vs GC | 2 | 0.68 (0.016) | 0.74 (0.006) | 0.70 (0.01) | 0.67 (0.04) | 0.70 (0.03) | 0.70 (0.01) |
1 p-values were computed by permutation testing as the fraction of permuted statistics that are at least as extreme as the test statistic obtained using the original class labels.