Table 4.
Performance of miRNAs selected by PCA-based FE with PCA-based LDA and SVM
| Number of | ||||||
|---|---|---|---|---|---|---|
| Diseases | Accuracy | Sens. | Spec. | miRNAs * | PCs + | Δ # |
| PCA-based LDA | ||||||
| AD | 0.886 | 0.917 | 0.818 | 22 | 16 | 2.5 |
| Carcinoma | 0.857 | 0.846 | 0.867 | 36 | 2 | 7 |
| CAD | 0.885 | 0.923 | 0.846 | 16 | 14 | 9 |
| NPC | 0.720 | 0.806 | 0.579 | 28 | 18 | 5 |
| HCC | 0.650 | 0.600 | 0.700 | 8 | 1 | 7 |
| BC | 1.000 | 1.000 | 1.000 | 18 | 13 | 6 |
| AML | 0.862 | 0.846 | 0.923 | 11 | 8 | 7 |
| Mean | 0.837 | 0.848 | 0.819 | |||
| Mean of previous study [23] | 0.784 | 0.750 | 0.800 | |||
| SVM | ||||||
| AD | 0.843 | 0.833 | 0.864 | 22 | ||
| Carcinoma | 0.786 | 0.807 | 0.767 | 36 | ||
| CAD | 0.807 | 0.615 | 1.000 | 16 | ||
| NPC | 0.720 | 0.774 | 0.632 | 28 | ||
| HCC | 0.770 | 0.550 | 0.850 | 8 | ||
| BC | 0.963 | 1.000 | 0.938 | 18 | ||
| AML | 0.969 | 1.000 | 0.846 | 11 | ||
| Mean | 0.837 | 0.797 | 0.842 | |||
*number of miRNAs selected by PCA-based FE, +optimal number of PCs estimated by LOOCV, #threshold value of PCA-based FE. Data from previous study [23] are also shown for comparison. AD, Alzheimer’s disease; CAD, coronary artery disease; NPC, nasopharyngeal carcinoma; HCC, hepatocellular carcinoma; BC, breast cancer; AML, acute myeloid leukemia; UDB, universal disease biomarker; SVM, support vector machine; LDA, linear discriminant analysis; PCA, principal component analysis.