Predicting performance comparison of the proposed method with four existing methods using PPI data to identify informative genes.

<p>For each experiment, the optimal combination of two thresholds was obtained using the approach mentioned above and was applied to an independent test using unlabeled samples. Bold font indicates the superior performer.</p><p>TSVM: <i>P</i> (the ratio of two class labels).</p><p>SVM: PolyKernel –C 250007–E 1.0, The complexity parameter C (1.0), epsilon (1.0E−12), filterType (Normalized training data).</p><p>Naïve Bayesian: No parameters.</p><p>Random Forest: numTrees (10), seed (1).