TABLE 2.
Performance of the best classifiers for eight cell types based on two classification algorithms.
| Cell type | Classification algorithm (Feature list) | Number of features | Weight F1 | MCC | ACC |
|---|---|---|---|---|---|
| Blood B cells | DT (MCFS feature list) | 70 | 0.712 | 0.627 | 0.71 |
| RF (LightGBM feature list) | 200 | 0.894 | 0.863 | 0.894 | |
| Blood T cells | DT (mRMR feature list) | 1,060 | 1.000 | 1.000 | 1.000 |
| RF (mRMR feature list) | 1,220 | 0.971 | 0.961 | 0.971 | |
| Nasal B cells | DT (MCFS feature list) | 1900 | 0.610 | 0.483 | 0.611 |
| RF (MCFS feature list) | 1,520 | 0.779 | 0.706 | 0.779 | |
| Nasal T cells | DT (LightGBM feature list) | 80 | 0.607 | 0.489 | 0.605 |
| RF (MCFS feature list) | 1,040 | 0.773 | 0.715 | 0.777 | |
| Nasal macrophages | DT (LightGBM feature list) | 70 | 0.731 | 0.64 | 0.727 |
| RF (LightGBM featue list) | 1760 | 0.870 | 0.825 | 0.870 | |
| Lung macrophages | DT (LightGBM feature list) | 100 | 0.733 | 0.660 | 0.733 |
| RF (LightGBM feature list) | 110 | 0.838 | 0.795 | 0.839 | |
| Alveolar epithelial cells | DT (mRMR feature list) | 1,470 | 1.000 | 1.000 | 1.000 |
| RF (mRMR feature list) | 1,660 | 0.873 | 0.838 | 0.873 | |
| Lung endothelial cells | DT (LightGBM feature list) | 60 | 0.753 | 0.678 | 0.750 |
| RF (LightGBM feature list) | 170 | 0.924 | 0.901 | 0.924 |