Skip to content

Protein-Ligand binding affinity prediction with Protein LLMs and Graph Attention Networks.

Notifications You must be signed in to change notification settings

d-f/binding-affinity

Repository files navigation

To determine the mean and standard deviation of dataset features:

python determine_affinity_stats.py -save_filepath /affinity/ds_stats.json

To partition the dataset into training, validation and test partitions:

python partition_dataset.py -data_folder /affinity/processed/ -save_folder /affinity/

To embed the dataset proteins with ProtBERT and dataset ligands with ChemBERTa:

python process_plapt_ds.py -stat_json_path /affinity/ds_stats.json -save_dir /affinity/processed/

To train a PLAPT model to predict binding affinity:

python train_plapt.py -data_folder /affinity/processed/ -csv_folder /affinity/csv/ -batch_size 64 -lr 1e-3 -num_epochs 64 -result_folder /affinity/models/ -model_save_name model_1.pth.tar -patience 5 -prot_hidden 512 -lig_hidden 512

About

Protein-Ligand binding affinity prediction with Protein LLMs and Graph Attention Networks.

Topics

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages