AI-antibodies

This is the repo with all the material for generating and analysing HADDOCK3 files for machine learning-modelled antibodies.

1. create the environment with conda

conda create -n aiabs python=3.10.6

2. install the package with pip

pip install .

3. run the setup for your benchmark

for each pdb in your benchmark set, you can run the following command:

aiabs ${PDB} --input_dir=benchmark_haddock_23_May_2023/${PDB} --output_dir=${path_to_output_dir} --act_act_path=./bin/generate-act-act.sh

4. analyse the results

aiabs-analysis --ref_folders=haddock_runs_caprieval

You can exclude some antibodies from the analysis

aiabs-analysis --ref_folders=haddock_runs_caprieval --exclude_pdbs=7seg,7kn4

for plotting the results you can access the script available in the analysis folder.

Notebooks

Additional plots for correlating DockQ vs confidence measures and against interface/paratope/epitope RMSD can be created using the notebooks in analysis/notebooks.