This README contains the commands needed to apply the Virtual Lab to nanobody design for one of the latest variants of SARS-CoV-2.
After following the general virtual_lab installation instructions, run the following to install nanobody-specific requirements:
conda activate virtual_lab
cd /path/to/virtual_lab
pip install -e .[nanobody-design]If there are any version incompatibility issues, please run pip install -r nanobody_design/requirements_nanobody_design_frozen.txt.
Additionally, create a separate virtual environment for running AlphaFold-Multimer via LocalColabFold by following these instructions: https://github.com/YoshitakaMo/localcolabfold (or by running install_localcolabfold.sh). Make sure the following versions are installed:
alphafold-colabfold==2.3.6
colabfold==1.5.5
Run the run_nanobody_design.ipynb notebook to have LLM agents create a nanobody design workflow. Each meeting with the LLM agents should only take a few minutes. The outputs of these notebooks are in the nanobody_design/discussions directory. The relevant ESM, AlphaFold-Multimer, and Rosetta scripts have been copied from those discussions to the nanobody_design/scripts folder. Additionally, the sequences for the relevant nanobodies selected by the LLM agents, along with SARS-CoV-2 spike RBD sequences, are in the nanobody_design/sequences directory.
Starting with four nanobodies designed to bind to the original Wuhan strain of SARS-CoV-2---Ty1, H11-D4, Nb21, and VHH-72---improve them by iteratively adding mutations and scoring them using ESM, AlphaFold-Multimer, and Rosetta for four rounds.
See commands in nanobody_design/workflow.md.