Table 2 Five-fold cross validation results on profile HMMs learned from all data and seven MDD-clustered subgroups.
Models | Number of positive data | Number of negative data | Sn | Sp | Acc | MCC |
|---|---|---|---|---|---|---|
Single HMM with all data | 410 | 410 | 68.8% | 70.7% | 69.8% | 0.395 |
HMM with OGT1 | 100 | 100 | 93.0% | 89.0% | 91.0% | 0.821 |
HMM with OGT2 | 105 | 105 | 83.8% | 71.4% | 77.6% | 0.557 |
HMM with OGT3 | 95 | 95 | 85.3% | 75.8% | 80.5% | 0.613 |
HMM with OGT4 | 39 | 39 | 71.8% | 74.4% | 73.1% | 0.462 |
HMM with OGT5 | 30 | 30 | 73.3% | 73.3% | 73.3% | 0.467 |
HMM with OGT6 | 19 | 19 | 78.9% | 73.7% | 76.3% | 0.527 |
HMM with OGT7 | 22 | 22 | 72.7% | 68.2% | 70.5% | 0.410 |
MDD-clustered HMMs (Combined 7 OGT HMMs) | 410 | 410 | 83.7% | 77.1% | 80.4% | 0.609 |
Two-layered model (7 HMMs + 1 SVM) | 410 | 410 | 85.4% | 84.1% | 84.7% | 0.695 |