Table 8.
Performance of a nine-protein signature derived from Parallel Regularised Regression Model Search.
| Training set | Test set | |
|---|---|---|
| ROC AUC | 90 (86–95) | 84 (73–94) |
| Sensitivity | 83 (75–90) | 86 (73–95) |
| Specificity | 82 (73–89) | 74 (58–86) |
| PPV | 84 (75–91) | 85 (70–94) |
| NPV | 82 (73–89) | 76 (62–87) |
| After adjustment of cut-off for increased sensitivity | ||
| Sensitivity | 90 (83–95) | 92 (80–98) |
| Specificity | 71 (61–80) | 71 (56–84) |
| PPV | 75 (66–82) | 75 (62–86) |
| NPV | 89 (80–95) | 90 (76–97) |
| After adjustment of cut-off for increased specificity | ||
| Sensitivity | 73 (63–82) | 75 (60–87) |
| Specificity | 90 (82–95) | 81 (67–91) |
| PPV | 87 (78–94) | 80 (65–91) |
| NPV | 78 (69–85) | 77 (63–88) |
This analysis was performed using data on the ILULU-TB cohort only (n = 249). The same patients and allocations were used as for the GDA and Random Forests analyses (Tables 4, 5). This signature comprised fibrinogen, alpha-2-M, CRP, MMP-9, transthyretin, complement FH, IFN-gamma, IP-10, and TNF-alpha.