Spaces:
Running
on
Zero
Running
on
Zero
title: Allmetal3d | |
emoji: π | |
colorFrom: blue | |
colorTo: purple | |
sdk: gradio | |
sdk_version: 4.20.1 | |
app_file: app.py | |
pinned: false | |
license: mit | |
short_description: AllMetal3D | |
# AllMetal3D | |
 | |
AllMetal3D is an improvement of the Metal3D model trained using a wider amount of ions (NA, K, MG, CA, ZN, FE, FE2, CU, CO, MN). | |
The model will predict location, identity and geometry for each site. | |
Location is very accurate and can be evaluated using the predicted probability. | |
Identity also works well with the ion being classed into majority labels (Alkali, MG, CA, ZN, Non zinc transition metal and NoMetal). NoMetal are likely false positive location predictions. | |
Geometry prediction works well to differentiate between octahedral and tetrahedral but not so well for the other classes. | |
## How to run | |
In webbrowser for single structures and visual processing use the gradio app. | |
``` | |
python app.py | |
``` | |
For running in batch mode use the commandline tool: | |
``` | |
python run_metalsitepred.py --pdb 2cba.pdb --cubefile test.cube --probefile 2cba_probe.pdb --batch_size 30 | |
``` | |