Code and datasets for an antibody binding ΔΔG protocol designed in Rosetta to improve simulation of antibody evolution. The protocol is that of Barlow et al. 2018 (GitHub), with only a few parameter changes.

To obtain ddGs for a saturation mutagenesis type procedure, there are three main files to pay attention to: predict_ddg.sh, ddgs_final.sh, and get_csv.sh.

For CU Boulder users, these are all set up for auto-submission of a job to Alpine on the amilan-ucb partition. A job will be submitted for each mutation position you enter. Please also remember to enter your email in the --mail-user= tag.

Make sure you have installed the Anaconda environment used for this project contained in env.yaml.

conda env create --file=env.yaml

Then ensure that you have either cloned the whole repository or have downloaded the python and/or XML files necessary for the shell scripts to run — ddG_backrub_og.xml, ddG_backrub.xml, calc_ddgs.py, analyze_ddGs.py, run_ddgs.py, and the shell scripts themselves.

harmonic_prerelax.py is also helpful if your structure files have not been relaxed into Rosetta yet. To do this, run:

python harmonic_prerelax.py PATH_TO_PDB

Now, to run the PyRosetta version of the script, use predict_ddg_py.sh. To run the RosettaScripts version, use predict_ddg.sh. I have listed the process for predict_ddg_py.sh here. Both take the same command line arguments, but the RosettaScripts version requires that you set the path to your RosettaScripts executable in run_ddgs.py. The PyRosetta version has the REF2015 score function as default, while the RosettaScripts version defaults to the Talaris score function. The PyRosetta version does not yet output PDBs like the RosettaScripts version.

Running the script:

./predict_ddg_py.sh -p POSITION -o OUTPUT_PATH -i INPUT_PDB_PATH -j JUMP_NUMBER -n NUM_TASKS

Position arguments must be listed in the form (CHAIN):(WT_AA)(PDB_INDEX), e.g. H:D32 or L:S100B with an insertion code. Multiple positions can be entered at once, and should be entered as a bash array with spaces between positions, e.g. (H:D32 L:S100B ...)

The jump number will be used to properly unbind the antibody from the antigen, and can be found using the find_jumps.ipynb notebook.

The NUM_TASKS argument is the number of parallel jobs to run. However, the job set up by the predict_ddg.sh script will only run one job at a time, so this argument must be 1. This may be changed later.

An example script that can be run using files in this repository is:

./predict_ddg_py.sh -p H:D32 -o ./example -i ./inputs/1DQJ_all.clean.pdb -j 1 -n 1

Please open an issue on GitHub if you have any questions or issues.