Table 3.
Fivefold cross-validation of the prediction accuracy of support vector machine (SVM) classifiers corresponding to different combinations of omics and clinical features.
| Inputs for the SVM Classifier | Prediction Accuracy (%) | Sensitivity (%) | Specificity (%) |
|---|---|---|---|
| Autoencoder model-generated nodes (#53) | 97.1% | 97.0% | 97.2% |
| Multiomics features (#250) 1 | 90.8% | 93.8% | 87.4% |
| Multiomics features + 3 clinical features 2 | 93.7% | 95.6% | 91.5% |
| Multiomics features + stage | 94.3% | 96.5% | 91.7% |
| Multiomics features + age | 90.7% | 93.6% | 87.4% |
| Multiomics features + Gleason Score | 90.6% | 93.2% | 87.6% |
1 Multiomics features included the top 100 differentially expressed genes, the top 100 differentially expressed methylation genes, and the top 50 differentially expressed miRNAs. 2 The three clinical features included were the Gleason score, age, and stage data.