Table 1 Diagnostic performance of the LDS model across training and validation cohorts
Cohort | AUC (95% CI) | Sensitivity (%, 95% CI) | Specificity (%, 95% CI) | Overall accuracy (%, 95% CI) |
|---|---|---|---|---|
Train | 0.897 (0.849–0.934) | 83.5 (75.8–89.0) | 80.4 (71.6–86.9) | 82.1 (76.5–86.5) |
GSE122063 | 0.866 (0.792–0.925) | 71.4 (58.5–81.6) | 88.6 (76.0–95.0) | 79.0 (70.0–85.8) |
GSE132903 | 0.826 (0.762–0.879) | 79.4 (70.3–86.2) | 67.3 (57.6–75.8) | 73.3 (66.7–79.0) |
GSE28146 | 0.812 (0.602–0.989) | 72.7 (51.8–86.8) | 75.0 (40.9–92.9) | 73.3 (55.6–85.8) |
GSE29378 | 0.780 (0.659–0.882) | 80.6 (63.7–90.8) | 56.2 (39.3–71.8) | 68.3 (56.0–78.4) |
GSE36980 | 0.763 (0.648–0.867) | 75.8 (59.0–87.2) | 68.1 (53.8–79.6) | 71.3 (60.5–80.0) |
GSE37263 | 0.906 (0.719–1.000) | 100.0 (67.6–100.0) | 75.0 (40.9–92.9) | 87.5 (64.0–96.5) |
GSE48350 | 0.729 (0.665–0.791) | 78.7 (68.6–86.3) | 52.6 (45.2–59.9) | 60.9 (54.7–66.7) |
Independent clinical plasma cohort | 0.772 (0.729–0.815) | 48.1 (42.3–54.1) | 96.7 (92.9–98.5) | 67.6 (63.1–71.7) |
