Table 6 |
|||
|
Comparison of our kernel-based integration approach with the ensemble approach |
|||
|
Outcome |
AUC (SE)*:MPT1/MG |
AUC (SE)*: ensemble approach |
p-value |
|
|
|||
|
WHEELER |
0.9269 (0.0425) |
0.9500 (0.0339) |
0.6160 |
|
pN-STAGE |
0.9870 (0.0135) |
0.9253 (0.0432) |
0.1422 |
|
CRM |
0.9630 (0.0344) |
0.7860 (0.0783) |
0.0384 |
|
GRADE |
0.9006 (0.0413) |
0.8567 (0.0521) |
0.3745 |
|
STAGE |
0.8528 (0.0550) |
0.8304 (0.0582) |
0.6836 |
|
METASTASIS |
0.9868 (0.0121) |
0.9452 (0.0309) |
0.1313 |
|
RECURRENCE |
0.7857 (0.0934) |
0.4545 (0.1352) |
0.0182 |
|
|
|||
|
*Area under the ROC curve (standard error) obtained with leave-one-out. †Comparison in AUC between the best models obtained with our strategy (MPT1 for rectal cancer, MG for prostate cancer) and the corresponding ensemble models based on the same number of features [46] |
|||
|
Daemen et al. Genome Medicine 2009 1:39 doi:10.1186/gm39 |
|||
