Inference docker for:

Boosting Convolutional Neural Networks' Protein Binding Site Prediction Capacity Using SE(3)-invariant transformers, Transfer Learning and Homology-based Augmentation

Daeseok Lee, Jeunghyun Byun, Bonggun Shin

arXiv

docker pull daeseoklee/bsp-inference

Model checkpoints (including those from the ablation study) are included in the docker image. Optionally, one can download them from huggingface. After downloading train_logs.tar and unzipping, one can perform inference by running src/inference/run.py without using docker.

# examples/ is a directory containing pdb files 
./predict-binding-site examples examples/out.csv

Optionally, one can perform inference for the datasets appearing in our paper (scPDB etc.). The datasets (preprocessed in HDF5 format) can be downloaded from huggingface. Then, one can run, for example,

./predict-binding-site scPDB_cache_partitioned/scPDB_cache_1.hdf5 scPDB_out_1.csv 

rm ./casestudy/out.csv
bash ./casestudy/inference.sh