NEW - You can now run NanoNet predictions using our webserver, https://bio3d.cs.huji.ac.il/nanonet/
NanoNet - a rapid nanobody modeling tool. for citation, please cite our paper: https://www.frontiersin.org/articles/10.3389/fimmu.2022.958584/full
How to run NanoNet from google Colaboratory:
1. Open the Colab notebook (NanoNet.ipynb, link above).
2. Select protein type (Nb/mAb heavy chain or TCR VB).
3. Select input type (sequence (String) or path to a fasta file)
4. Provide a Nb sequence/fasta (NanoNet will preduce a model for each entry in the fasta file).
5. Select whether or not you want to reconstruct the side chains using modeller (requires license - https://salilab.org/modeller/).
6. Press the 'Run all' option.
How to run NanoNet locally:
1. Clone the git repository : git clone "https://github.com/dina-lab3D/NanoNet"
2. Make sure you have the following libraries installed in your environment:
- numpy
- tensorflow (2.4.0 or higher)
- Bio
- modeller (optional, only if you want to reconstruct the side chains using modeller, requires license - https://salilab.org/modeller/)
3. If you want to reconstruct the side chains using Scwrl4, make sure you have it installed on your computer (requires license - http://dunbrack.fccc.edu/SCWRL3.php/).
4. Run the following command (with python 3):
python NanoNet.py <fasta file path>
this will produce a backbone + cb pdb named '<record name>_nanonet_backbone_cb.pdb' for each record in the fasta file.
options:
-s : write all the models into a single PDB file, separated with MODEL and ENDMDL (reduces running time when predicting many structures), default is False.
-o <output directory> : path to a directory to put the generated models in, default is './NanoNetResults'
-m : run side chains reconstruction using modeller, default is False. Output it to a pdb file named '<record name>_nanonet_full_relaxed.pdb'
-c <path to Scwrl4 executable>: run side chains reconstruction using scwrl, default is False. Output it to a pdb file named '<record name>_nanonet_full.pdb'
-t : use this parameter for TCR V-beta modeling, default is False
Running times for 1,000 structures on a single standard CPU:
only backbone + Cb - less than 15 seconds (For better preformance use GPU and cuda). backbone + SCWRL - about 20 minutes. backbone + Modeller - about 80 minutes.