Table 1. Area under the ROC curve (AUC), specificity and sensitivity of the best performing parenclitic and RDLG models for early (>34.5 months to diagnosis) and late (<14.5 months to diagnosis) groups.
| Group | Type | Model | AUC | Specificity | Sensitivity |
|---|---|---|---|---|---|
| Early | Parenclitic | -7.44 (Alpha-centrality) | 0.624 | 1 | 0.189 ± 0.124 |
| -0.16 (Pill use) | |||||
| RDLG | -2.09 (GPC1) | 0.663 | 1 | 0.163 ± 0.087 | |
| Late | Parenclitic | 0.07 (Connections to MUC16) | 0.904 | 1 | 0.803 ± 0.077 |
| RDLG | 2.14 (MUC16) | 0.904 | 1 | 0.767 ± 0.082 |
Models were generated using all combinations of 1, 2 or 3 markers and following cross-validation the best models were selected based initially on sensitivity and secondly on AUC. Specificity was set to be >98% and the maximum is reported for the given sensitivity. Models were generated by logistic regression of either the parenclitic topologies or raw data and model coefficients are provided.