Geometric Deep Learning for Molecular Interactions

This repository contains the source code and documentation related with Sofia Guerreiro's research on Geometric Deep Learning for molecules binding.

MSc Thesis title (Instituto Superior Técnico), Nov. 2023 - Sofia Guerreiro:

"Predicting Protein-Ligand Binding Affinity using Graph Neural Networks"

Running the code

This code was developed using Python version 3.10. It is recommended to use the same or a compatible version of Python.

If you don't already have Python 3.10 installed, you can download it from the official Python website.

Mlflow

We are using mlflow to keep track of our experiments. We have our own remote server to which we can log everything from any computer which was created following the instructions in this tutorial.

This is a nice feature to have but it is not required in any way to run our code. Everything will be logged locally by default. That is, all experiments will be logged to the folder mlruns created in the directory from which the script is launched. To then look at the experiments in the browser we just need to run the command mlflow ui.

Installing everything

The next step is to actually clone the repository using:

git clone git@github.com:inductiva/molecules-binding.git

The very next step is to create a virtual environment. This will solve any clashes with the library versions used here and anything else that might be installed in your own system:

python3 -m venv .env
source .env/bin/activate

After creating and activating the virtual environment we can install all the requirements of the project using:

pip install --upgrade pip
pip install -r requirements.txt

Next, because molecules-binding is actually packaged we can install it using:

pip install -e .

Processing the dataset

We included a small example dataset, in the directory example_dataset. For using a real dataset, you can download PDBBind. In this project, the majority of the experiments included PDBBind general set 2016. To process the dataset and store it, first create a directory where you want to keep stored datasets (e.g., /datasetsprocessed/), and then run the script process_dataset_interaction.py. For instance,

python scripts/process_dataset_interaction.py --affinity_dir=example_dataset/index/INDEX_general_PL_data.2020 --data_dir=example_dataset/ --path_dataset=../datasetsprocessed/example_data_processed --threshold=8 --which_file_ligand="mol2" --not_include_test_set=True --separate_edges=False

Training the model

To train a model is simply a matter of running the script train_graphnet_lightning.py:

python scripts/train_graphnet_lightning.py --path_dataset=../datasetsprocessed/example_data_processed --dropout_rate=0.1 --max_epochs=2500 --use_gpu=True --batch_size=3 --num_hidden_linear=256,256 --train_split=0.9 --learning_rate=0.0001 --weight_decay=0.0001 --use_batch_norm=True --comment="running final architecture" --embedding_layers=128,128 --use_message_passing=True --which_gnn_model=NodeEdgeGNN --size_processing_steps=128 --early_stopping_patience=500 --num_processing_steps=3 --splitting_seed=24 --save_model=False --mlflow_server_uri=<Your mlflow server uri>

You can choose other flags, with different parameters.

Evaluating the model on a test set

If you choose to save the model, you can later evaluate the performance on other test sets. You can create a directory to store the results (e.g., /results/) and run the script evaluate_model.py. For instance,

python scripts/evaluate_model.py --path_dataset=<Path to the processed dataset to test> --mlflow_server_uri=<your mlflow server ui> --results_dir=../results --run_id=<The run ID where the model was trained>