About

This code repository accompanies the paper

Inferring protein fitness landscapes from laboratory evolution experiments
Sameer D’Costa, Emily C. Hinds, Chase R. Freschlin, Hyebin Song, Philip A. Romero

• Biorxiv link doi

Data for this repository can be downloaded from NCBI Sequence Read Archive Accession: PRJNA923701

Downloading this repository

Use git to clone this repository from github.com

Downloading raw sequencing data (Optional)

Raw sequencing data can be downloaded using the convenience script in the preprocessing directory.

(cd preprocessing; ./download_data.sh)

Repository Installation

Use conda to install dependencies. The main dependencies are

• pytorch version 1.6
• python version 3.8

There are several other dependencies like numpy, matplotlib etc that are not fixed and the latest versions should be fine.

# create conda `dhfr_analysis` environment
make install_conda

conda activate dhfr_analysis

# (Recommended) Adding python kernel in this conda environment to jupyter
#   notebook list of kernels
python3 -m ipykernel install --user

# make c++ extensions to python code
make ext

The exact the packages that this repository was tested on is saved at environment.explicit.txt

Running test script

# activate conda environment (if not already active)
conda activate dhfr_analysis

#run a test script 
# (This will print out a list of allowed arguments)
./runmodel.sh -h

The output should be

usage: model.py [-h] [-l LEARNING_RATE] [-j] [-n NUM_EPOCHS] [-m MODEL_NAME]
                [-v LOG_LEVEL] [-o OUTPUT_DIRECTORY] [-t NT_TRANS_MAT_CSV]
                [-L PROTEIN_LENGTH] [--lam_main LAM_MAIN] [--lam_int LAM_INT]
                [-i] [--margin_penalty MARGIN_PENALTY]

optional arguments:
  -h, --help            show this help message and exit
  -l LEARNING_RATE, --learning_rate LEARNING_RATE
                        Learning rate for ADAM optimizer
  -j, --disable_joblib  Disable saving of joblib cache
  -n NUM_EPOCHS, --num_epochs NUM_EPOCHS
                        Number of training steps
  -m MODEL_NAME, --model_name MODEL_NAME
                        Model Name
  -v LOG_LEVEL, --log_level LOG_LEVEL
                        Log level for debugging messages
  -o OUTPUT_DIRECTORY, --output_directory OUTPUT_DIRECTORY
                        Output directory for model params and figs (Path
                        Relative to the scripts directory)
  -t NT_TRANS_MAT_CSV, --nt_trans_mat_csv NT_TRANS_MAT_CSV
                        Nucleotide Transition Matrix pickle file (Path
                        Relative to the scripts directory)
  -L PROTEIN_LENGTH, --protein_length PROTEIN_LENGTH
                        Change protein length for debugging
  --lam_main LAM_MAIN   Regularization for main effects parameters
  --lam_int LAM_INT     Regularization for interaction effects parameters
  -i, --include_main_params
                        Include Main Parameters
  --margin_penalty MARGIN_PENALTY
                        Penalty for margin loss

Reinstallation

These are the steps to reinstall the latest version of the code for the second time after any changes.

# Make sure you are in the top directory of this repository

# pull the latest version of the code
git checkout master
git pull

# This deletes the `dhfr_analysis` conda environment without asking.  
make reinstall_conda

# clean up extra files
make deepclean
make ext