Binding affinity prediction for protein-ligand complex using deep attention mechanism based on intermolecular interactions

Abstract

Background: Accurate prediction of protein-ligand binding affinity is important for lowering the overall cost of drug discovery in structure-based drug design. For accurate predictions, many classical scoring functions and machine learning-based methods have been developed. However, these techniques tend to have limitations, mainly resulting from a lack of sufficient energy terms to describe the complex interactions between proteins and ligands. Recent deep-learning techniques can potentially solve this problem. However, the search for more efficient and appropriate deep-learning architectures and methods to represent protein-ligand complex is ongoing.

Results: In this study, we proposed a deep-neural network model to improve the prediction accuracy of protein-ligand complex binding affinity. The proposed model has two important features, descriptor embeddings with information on the local structures of a protein-ligand complex and an attention mechanism to highlight important descriptors for binding affinity prediction. The proposed model performed better than existing binding affinity prediction models on most benchmark datasets.

Conclusions: We confirmed that an attention mechanism can capture the binding sites in a protein-ligand complex to improve prediction performance. Our code is available at https://github.com/Blue1993/BAPA .

Keywords: Attention mechanism; Binding affinity; Protein–ligand complex; Structure-based drug design.

Fig. 1

Overview of BAPA experiments. First, we collected training data from the PDBbind database. Second, we constructed a deep learning model that captures local structure patterns that can help predict binding affinity by using convolutional and attention layers. Third, the optimal parameters were found using the validation dataset. Finally, the performance of the model was evaluated using an external test dataset. The numbers (#) of protein–ligand complexes are summarized in each step

Fig. 2

Average ranking comparison results for test datasets. a Protein structure generalization test results with lowest pairwise-chains TM-score. b Ligand structure generalization test results

Fig. 3

Visualization of interactions sites with high attention score. a 1EBY complex, b 3DD0 complex. The green dash lines and the brick-red spoked arcs indicate hydrogen bonds and hydrophobic contacts between the two atoms, respectively. Interactions sites captured by BAPA are highlighted in yellow

Fig. 4

Example of descriptor. There are two atom pairs between which the distance less than 12 $\mathrm{\AA }$, and from which two descriptors can be identified. The frequencies in the protein–ligand complex are counted for all the unique descriptors in the training dataset, with the results being d. Note that OH is regarded as O, and atoms that are not specified are carbon C