Table 2. Accuracy of class prediction based on the expression profile of nine MMC responsive genes.
| Class | Expression data sourcea | Mean percent of correct classification | ||||
| DLDA | 1-NN | NC | SVM | CCP | ||
| BRCA1 vs BRCA2 vs BRCAX | Pools (Microarray) | 100% | 67% | 89% | – | – |
| Pools (QRT-PCR) | 56% | 67% | 78% | – | – | |
| Virtual Pools (QRT-PCR) | 78% | 89% | 67% | – | – | |
| Samples (QRT-PCR) | 48% | 52% | 59% | – | – | |
| BRCA1 vs BRCA2 | Pools (Microarray) | 83% | 100% | 100% | 100% | 100% |
| Pools (QRT-PCR) | 100% | 100% | 100% | 83% | 100% | |
| Virtual Pools (QRT-PCR) | 100% | 100% | 100% | 100% | 100% | |
| Samples (QRT-PCR) | 83% | 83% | 83% | 83% | 83% | |
| BRCA1 vs BRCAX | Pools (Microarray) | 100% | 100% | 100% | 100% | 100% |
| Pools (QRT-PCR) | 67% | 83% | 67% | 83% | 67% | |
| Virtual Pools (QRT-PCR) | 83% | 100% | 83% | 100% | 83% | |
| Samples (QRT-PCR) | 56% | 56% | 67% | 56% | 56% | |
| BRCA2 vs BRCAX | Pools (Microarray) | 100% | 50% | 83% | 83% | 100% |
| Pools (QRT-PCR) | 67% | 67% | 83% | 50% | 67% | |
| Virtual Pools (QRT-PCR) | 33% | 83% | 67% | 67% | 83% | |
| Samples (QRT-PCR) | 44% | 72% | 72% | 61% | 72% | |
a Pools, n = 9; Samples, n = 27. Abbreviations: CCP, Compound covariate predictor; DLDA, Diagonal Linear Discriminant Analysis; NC, Nearest Centroid; 1-NN, Nearest Neighbour; SVM, support vector machine.