PLAPT: Protein-Ligand Affinity Prediction Using Pretrained Transformers

PLAPT is a state-of-the-art tool for predicting protein-ligand binding affinity, crucial for accelerating drug discovery processes. Our model leverages transfer learning from pretrained transformers like ProtBERT and ChemBERTa to achieve high accuracy while requiring minimal computational resources.

Key Features

• Efficient Processing: Extremely lightweight prediction module allows for incredibly high throughput affinity prediction with cached embeddings.
• Transfer Learning: Uses pretrained models to extract rich protein and molecule features.
• Versatile Usage: Uses just 1D protein and ligand sequences as strings for input. Has a command-line interface and Python API for easy integration into various workflows.
• High Accuracy: Achieves top performance on benchmark datasets.

Read our preprint

Model Architecture

PLAPT uses a novel branching neural network architecture that efficiently integrates features from protein and ligand encoders to estimate binding affinities:

This architecture allows PLAPT to process complex molecular information effectively and highly efficiently when coupled with caching.

Quick Start

Installation

For extra assistance installing please see our Bindwell AI

1. Clone the repository:

   git clone https://github.com/trrt-good/WELP-PLAPT.git
   cd WELP-PLAPT

2. Choose one of the following installation methods:

   Option A: Using Conda (Recommended)

   conda env create -f environment.yml
   conda activate plapt

   Option B: Using Python Virtual Environment

   python3 -m venv env

   For macos or linux, run:

   source env/bin/activate

   For windows:

   env\Scripts\activate

   Then install dependencies:

   pip3 install -r requirements.txt

Using PLAPT

PLAPT can be used via command line or integrated into Python scripts.

Command Line Interface

Predict affinity for a single protein and multiple ligands:

python3 plapt_cli.py -p "SEQUENCE" -m "SMILES1" "SMILES2" "SMILES3"

Predict affinities for multiple protein-ligand pairs:

python3 plapt_cli.py -p "SEQUENCE1" "SEQUENCE2" -m "SMILES1" "SMILES2"

Use files for input:

python3 plapt_cli.py -p proteins.txt -m molecules.txt

Save results to a file:

python3 plapt_cli.py -p "SEQUENCE" -m "SMILES1" "SMILES2" -o results.json

Python Integration

from plapt import Plapt

plapt = Plapt()

# Predict affinity for a single protein and multiple ligands
protein = "MKTVRQERLKSIVRILERSKEPVSGAQLAEELSVSRQVIVQDIAYLRSLGYNIVATPRGYVLAGG"
molecules = ["CC1=CC=C(C=C1)C2=CC(=NN2C3=CC=C(C=C3)S(=O)(=O)N)C(F)(F)F", 
             "COC1=CC=C(C=C1)C2=CC(=NN2C3=CC=C(C=C3)S(=O)(=O)N)C(F)(F)F"]

results = plapt.score_candidates(protein, molecules)
print(results)

# Predict affinities for multiple protein-ligand pairs
proteins = ["SEQUENCE1", "SEQUENCE2"]
molecules = ["SMILES1", "SMILES2"]

results = plapt.predict_affinity(proteins, molecules)
print(results)

Docker

You can also run the model using Docker:

1. Build the Docker image locally:

   docker build -t plapt .

   Or pull the pre-built image from Docker Hub:

   docker pull cford38/plapt:latest

2. Run the Docker container:

   docker run --gpus all --rm --name plapt -it plapt /bin/bash
   # docker run --gpus all --rm --name plapt -it cford38/plapt:latest /bin/bash

Note

This image does not include all of the model weights, which will be downloaded the first time you run PLAPT inside in the container.

Used by

PLAPT has been used in the following research:

1. López-Cortés, A., Cabrera-Andrade, A., Echeverría-Garcés, G. et al. Unraveling druggable cancer-driving proteins and targeted drugs using artificial intelligence and multi-omics analyses. Sci Rep 14, 19359 (2024). https://doi.org/10.1038/s41598-024-68565-7

If you've used PLAPT in your research, please let us know!

Citation

If you use PLAPT in your research, please cite our paper:

@misc{rose2023plapt,
  title={PLAPT: Protein-Ligand Binding Affinity Prediction Using Pretrained Transformers},
  author={Tyler Rose, Nicolò Monti, Navvye Anand, Tianyu Shen},
  journal={bioRxiv},
  year={2023},
  url={https://www.biorxiv.org/content/10.1101/2024.02.08.575577v3},
  doi={10.1101/2024.02.08.575577}
}

License

This project is licensed under the MIT License - see the LICENSE file for details.