Binder design loop

Code to perform binder design loop on CD20 (lymphoma marker). The main script, run_design_loop.py, uses colabfold (i.e., AF2 Multimer) and ProteinMPNN sequentially and iteratively to optimize a peptide binder to the target protein.

Installation

You should follow the instructions from localcolabfold to install that package. You will need to include the colabfold path in your PATH variable in ~/.bashrc or edit the colabfold_template.sh file to include the full path to colabfold_batch.

Then, follow the instructions in ProteinMPNN to create a conda environment that can run those codes and clone the GitHub repo in your machine. You will need to provide the path to the Python executable from that environment and the paths to the ProteinMPNN codes in proteinMPNN_template.sh.

You will also need to have a access to a Python environment with numpy, mdtraj, and natsort installed to run the script run_design_loop.py. This environment can be the same as the one that runs ProteinMPNN.

Running the code

To run the code, first open run_design_loop.py. Edit line 153 according to your desired output path and total number of runs.

def main():
    run_loop(outpath="./loop_oct_2/", initial_round=0, total_rounds=10)

Notice that you can restart from a round larger than 0 if you already ran some of them, but the loop will have no memory of the best models produced in past rounds.

You can control the chains and residue positions (indexed from 1, not 0!) that will be designed by editing lines 137-138:

run_proteinMPNN(
            template_script="proteinMPNN_template.sh", 
            pdb_path=protein_mpnn_output, 
            chains_to_design="C", # Modify as needed
            positions_to_design="2 3 4 6 8 9 10 11 12 13 14 15 16 17 18 19 20 21 25 26 27 28 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 72 74 75 77 79 80", # Modify as needed
            output_path=protein_mpnn_output, 
            seqs_per_target=30
        )

To start a loop, create the folder where the output will be saved. Then, in that directory, create a folder round_0. This folder must contain a fasta file named designed_sequences_round_0.fa to start the loop. If you have a PDB file with the initial guess for the binder and want to start from there, follow these steps to generate the fasta file:

1. Create a copy of proteinMPNN_template.sh (e.g., proteinMPNN_init.sh)
2. Edit proteinMPNN_init.sh: change "{INPUT_PDB_PATH}" with the path to the PDB file and change "{OUTPUT_PATH}" with the path to round_0.
3. Run proteinMPNN_init.sh. Check that ProteinMPNN created the output (.fa file) in the directory round_0/seqs.
4. Run the following code in an interactive Python session:

>>>from run_design_loop import prepare_sequence_to_structure # May need to change import depending on where you stored the code
>>>prepare_sequence_to_structure("{output_path}/round_0", "{output_path}/round_0", 0) # Change output path here

After the last step, the file round_0/designed_sequences_round_0.fa should be available.

Now, you can run the desing loop:

$ python run_desing_loop.py

Output

The output is organized as follows:

output_directory
	|
	round_{r}
		|_colabfold_results # Prdicted structures, scores, log file, etc.
		|_seqs # Sequences designed by ProteinMPNN
		|_designed_sequences_round_{r}.fa # Sequences from ProteinMPNN formatted for colabfold
		|_Other files from ProteinMPNN...

Known issues

For some unknown cause, colabfold_batch may produce segmentation faults after some rounds.