A wrapper around proteinMPNN to automate sequence generation and allow specifying fixed residues using regex (and ranges). See the included shell scripts for examples of how to use it.
- Clone proteinMPNN and follow their installation instructions.
- Clone this repository.
- Activate your conda environment associated with proteinMPNN.
- Run
better_mpnn.py
with the desired parameters.
The wrapper has the following usage:
usage: better_mpnn.py [-h] [--input_pattern INPUT_PATTERN] [--design_chains DESIGN_CHAINS] [--fix_pattern FIX_PATTERN] [--fix_ranges FIX_RANGES] [--num_outputs NUM_OUTPUTS]
[--sampling_temperature SAMPLING_TEMPERATURE] [--mpnn_path MPNN_PATH]
input_dir output_dir temp_dir
Run proteinMPNN on a set of input files
positional arguments:
input_dir The directory containing the input files
output_dir The directory to save the output files
temp_dir The directory to save temporary files
options:
-h, --help show this help message and exit
--input_pattern INPUT_PATTERN
The pattern to match for input files
--design_chains DESIGN_CHAINS
The chains to design
--fix_pattern FIX_PATTERN
The pattern to match for fixed residues
--fix_ranges FIX_RANGES
Ranges of residues to fix in the format "chain:res1-res2,chain:res3-res4"
--num_outputs NUM_OUTPUTS
The number of outputs per input file to generate
--sampling_temperature SAMPLING_TEMPERATURE
The "predictability" of the model
--mpnn_path MPNN_PATH
The path to the proteinMPNN directory
Running the script without any arguments will test some of the code, which may be useful if you want to extend this wrapper.