Table 2. Comparison of KNN to logistic regression.
| End point | Classifier | AUC | Common parametersa | MCC | Common parametersa | |||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|
| CV | P-value | Rank method | N | K | CV | P-value | Rank method | N | K | Threshold | ||
| Breast cancer: pathological complete response | KNN | 0.750 | 0.0005 | FC&(P<0.05) | 14 | 36 | 0.361 | 0.0037 | FC&(P<0.05) | 14 | 36 | 0.40 |
| LR | 0.708 | FC&(P<0.05) | 4 | NA | 0.247 | FC&(P<0.05) | 4 | NA | 0.23 | |||
| Breast cancer: estrogen receptor status | KNN | 0.952 | 0.3654 | FC&(P<0.05) | 9 | 25 | 0.847 | 0.4692 | P&(FC>1.5) | 5 | 15 | 0.70 |
| LR | 0.956 | FC&(P<0.05) | 5 | NA | 0.840 | FC&(P<0.05) | 4 | NA | 0.51 | |||
| Multiple myeloma: overall survival | KNN | 0.553 | 0.4390 | FC&(P<0.05) | 11 | 4 | 0.084 | 0.7561 | FC&(P<0.05) | 14 | 85 | 0.32 |
| LR | 0.564 | FC&(P<0.05) | 11 | NA | 0.092 | FC&(P<0.05) | 10 | NA | 0.53 | |||
| Multiple myeloma: event-free Survival | KNN | 0.636 | 0.0506 | P&(FC>1.5) | 15 | 15 | 0.245 | 0.0027 | P&(FC>1.5) | 16 | 39 | 0.40 |
| LR | 0.652 | P&(FC>1.5) | 10 | NA | 0.208 | FC&(P<0.05) | 11 | NA | 0.48 | |||
| Multiple myeloma: positive control | KNN | 0.962 | 0.0001 | FC&(P<0.05) | 13 | 18 | 0.834 | 0.4083 | FC&(P<0.05) | 7 | 152 | 0.49 |
| LR | 0.968 | FC&(P<0.05) | 5 | NA | 0.841 | FC&(P<0.05) | 5 | NA | 0.55 | |||
| Multiple myeloma: negative control | KNN | 0.527 | 0.7992 | P&(FC>1.5) | 10 | 8 | 0.045 | 0.3761 | P&(FC>1.5) | 9 | 8 | 0.31 |
| LR | 0.525 | FC&(P<0.05) | 10 | NA | 0.026 | FC&(P<0.05) | 12 | NA | 0.34 | |||
| Neuroblastoma: overall survival | KNN | 0.831 | 0.0001 | FC&(P<0.05) | 14 | 48 | 0.380 | 0.0000 | FC&(P<0.05) | 12 | 71 | 0.18 |
| LR | 0.768 | FC&(P<0.05) | 6 | NA | 0.262 | FC&(P<0.05) | 8 | NA | 0.31 | |||
| Neuroblastoma: event-free survival | KNN | 0.857 | 0.9658 | FC&(P<0.05) | 16 | 45 | 0.524 | 0.0673 | FC&(P<0.05) | 15 | 103 | 0.19 |
| LR | 0.857 | P&(FC>1.5) | 7 | NA | 0.499 | P&(FC>1.5) | 7 | NA | 0.20 | |||
| Neuroblastoma: positive control | KNN | 0.973 | 0.2942 | SAM | 4 | 40 | 0.909 | 0.1387 | SAM | 5 | 4 | 0.63 |
| LR | 0.970 | SAM | 4 | NA | 0.922 | SAM | 2 | NA | 0.29 | |||
| Neuroblastoma: negative control | KNN | 0.493 | 0.8727 | P&(FC>1.5) | 10 | 1 | -0.019 | 0.0636 | SAM | 9 | 26 | 0.40 |
| LR | 0.491 | FC&(P<0.05) | 9 | NA | 0.009 | FC&(P<0.05) | 8 | NA | 0.60 | |||
Abbreviations: AUC, area under the receiver operating characteristic curve; CV, cross-validation; FC, fold change; KNN, k-nearest neighbor; LR, logistic regression; MCC, Matthews correlation coefficient; SAM, significance analysis of microarrays.
Bold values indicate a P-value less than 0.005.
a
Mode of rank method and median of N, K and threshold.