Table 4 Recall, precision and F-measure for the proposed automated particle picking pipeline evaluated on EMPIAR-10017 (\(\beta\)-galactosidase) versus other state-of-the-art pickers
From: DRPnet: automated particle picking in cryo-electron micrographs using deep regression
Model training | Program | Version | # Picked particles | Recall (%) | Precision (%) | F-measure (%) |
|---|---|---|---|---|---|---|
Pretrained | DRPnet | 1 | 49,604 | 87.7 | 71.1 | 78.5 |
RELION | 3.0 | 49,855 | 73.4 | 59.9 | 65.9 | |
WARP | 1.0.9 | 46,438 | 83.9 | 73.6 | 78.4 | |
TOPAZ | cryoSPARC 2.14.2 | 51,749 | 83.4 | 64.9 | 73.0 | |
crYOLO(*) | 1.7.5 | 49,329 | 90.9 | 74.8 | 82.1 | |
DeepPicker | 1 | 40,880 | 26.5 | 26.4 | 26.5 | |
Trained from scratch with TRPV1 (EMPIAR-10005) | DRPnet | 1 | 49,604 | 87.7 | 71.1 | 78.5 |
RELION | 3.0 | 49,855 | 73.4 | 59.9 | 65.9 | |
WARP | 1.0.9 | 49,209 | 85.0 | 69.7 | 76.6 | |
TOPAZ | cryoSPARC 2.14.2 | 48,208 | 72.3 | 60.7 | 66.1 | |
crYOLO | 1.7.5 | 49,742 | 59.2 | 48.2 | 53.1 | |
DeepPicker | 1 | 40,205 | 23.3 | 23.6 | 23.5 |