CAPLA

— improved prediction of protein-ligand binding affinity by a deep learning approach based on a cross-attention mechanism.

Accurate and rapid prediction of protein-ligand affinity is a great challenge currently encountered in drug discovery. Recent advances have manifested a promising alternative in applying deep learning-based computational approaches for accurately quantifying protein-ligand affinity. The structure complementarity between protein-binding pockets and ligands determines the binding strength between a protein and a ligand, but most of existing deep learning approaches usually extracted the features of pockets and ligands by two detached modules. In this work, a new deep learning approach based on the cross-attention mechanism named CAPLA was developed for improved prediction of protein-ligand binding affinity based on sequence-level information of both protein and ligand alone. Specifically, CAPLA employs the cross-attention mechanism to crossly attend protein-binding pocket and ligand features, and further employs the dilated convolution to capture multiscale long-range contextual features. We evaluated the performance of our proposed CAPLA on multitudinous benchmarking experiments on protein-ligand binding affinity prediction, demonstrating the superior performance of CAPLA over state-of-the-art baseline approaches. Moreover, we provided the interpretability for CAPLA to uncover critical functional residues that contribute most to the protein-ligand binding affinity through the analysis of the attention scores generated by the cross-attention mechanism. Consequently, these results indicated that CAPLA is an effective approach for protein-ligand affinity prediction and may contribute to useful guidance for further drug development.

This source code was tested on the basic environment with conda==4.8.2 and cuda==11.3

Environment Reproduce

• In order to get CAPLA, you need to clone this repo:

  git clone [email protected]:lennylv/CAPLA.git
  cd CAPLA
  

• Unzip the "SourceCode.zip", "envs.zip" files into the current directory, and create environment using files provided in ./envs directory

  unzip SourceCode.zip
  unzip envs.zip
  cd envs
  conda env create -f capla_conda.yaml
  pip install -r capla_pip.txt
  

• Install the apex in the environment (the package is provided in ./envs/apex)

  cd apex
  python setup.py install
  

Run & Test

• Train your own model

  cd SourceCode/CAPLA/src
  python main.py
  

• Show the result in paper & Test your own dataset

  cd SourceCode/CAPLA/src
  python test.py testset (e.g., Test2016_290, Test2016_262)
  

contact

Zhi Jin : [email protected]

Tingfang Wu: [email protected]