Table 1. Overall comparison of classification performance among different techniques.
Disease group | Techniques | Precision (p) | Recall (r) | F-measure (F) |
Cardiovascular | PUDI | 82.0% | 80.3% | 80.4% |
ProDiGe | 54.3% | 96.3% | 69.3% | |
Smalter's method | 75.4% | 67.6% | 70.6% | |
Xu's method | 72.1% | 60.0% | 65.4% | |
EPU | 85.2% | 81.0% | 84.1% | |
Endocrine | PUDI | 83.6% | 75.3% | 79.2% |
ProDiGe | 57.3% | 87.7% | 69.3% | |
Smalter's method | 76.4% | 58.8% | 66.5% | |
Xu's method | 75.4% | 62.0% | 68.0% | |
EPU | 88.1% | 87.7% | 87.9% | |
Neurological | PUDI | 70.3% | 80.1% | 74.9% |
ProDiGe | 63.1% | 74.0% | 68.1% | |
Smalter's method | 60.6% | 65.9% | 63.1% | |
Xu's method | 59.7% | 66.7% | 63.0% | |
EPU | 78.2% | 80.4% | 78.6% | |
Metabolic | PUDI | 80.1% | 84.8% | 82.4% |
ProDiGe | 58.7% | 84.5% | 69.3% | |
Smalter's method | 59.1% | 84.7% | 69.6% | |
Xu's method | 65.6% | 78.3% | 71.4% | |
EPU | 83.3% | 93.9% | 90.9% | |
Ophthalmological | PUDI | 71.6% | 78.5% | 74.9% |
ProDiGe | 58.3% | 77.7% | 66.6% | |
Smalter's method | 56.7% | 77.8% | 65.5% | |
Xu's method | 64.2% | 71.3% | 67.4% | |
EPU | 89.3% | 81.0% | 84.7% | |
Cancer | PUDI | 76.3% | 80.0% | 78.0% |
ProDiGe | 71.1% | 79.8% | 75.3% | |
Smalter's method | 73.8% | 79.0% | 76.3% | |
Xu's method | 71.0% | 79.7% | 75.1% | |
EPU | 81.2% | 84.5% | 82.6% | |
Average performance | PUDI | 77.3% | 79.8% | 78.3% |
ProDiGe | 60.5% | 83.3% | 69.7% | |
Smalter's method | 67.0% | 72.3% | 68.6% | |
Xu's method | 68.0% | 69.7% | 68.4% | |
EPU | 84.2% | 84.8% | 84.8% |
PUDI is a SVM-based approach that partitions unlabeled genes into multiple levels with different associations to confirmed disease genes. ProDiGe is a bagging method that iteratively chooses random subsets from unlabeled subset and trains multiple classifiers. Smalter's method integrates multiple biological features, such as topological features, sequence-derived features, evolutionary age features. Xu's method employs the KNN classifier to predict disease genes.