Table 3.
Predictive power of the SVM classifiers
Set of input features to SVM classifier | Positive predictive value |
All graph theoretic protein characteristics | 0.08 |
All graph theoretic protein characteristics + 2Hop SS | 0.50 |
All graph theoretic protein characteristics + 2Hop SP | 0.15 |
All graph theoretic protein characteristics + 2Hop SS + 2Hop SP | 0.14 |
Positive predictive values for classifiers trained with and without 2Hop characteristics at a threshold level of 0.999.