Drug-target affinity prediction with extended graph learning-convolutional networks

Abstract

Background: High-performance computing plays a pivotal role in computer-aided drug design, a field that holds significant promise in pharmaceutical research. The prediction of drug-target affinity (DTA) is a crucial stage in this process, potentially accelerating drug development through rapid and extensive preliminary compound screening, while also minimizing resource utilization and costs. Recently, the incorporation of deep learning into DTA prediction and the enhancement of its accuracy have emerged as key areas of interest in the research community. Drugs and targets can be characterized through various methods, including structure-based, sequence-based, and graph-based representations. Despite the progress in structure and sequence-based techniques, they tend to provide limited feature information. Conversely, graph-based approaches have risen to prominence, attracting considerable attention for their comprehensive data representation capabilities. Recent studies have focused on constructing protein and drug molecular graphs using sequences and SMILES, subsequently deriving representations through graph neural networks. However, these graph-based approaches are limited by the use of a fixed adjacent matrix of protein and drug molecular graphs for graph convolution. This limitation restricts the learning of comprehensive feature representations from intricate compound and protein structures, consequently impeding the full potential of graph-based feature representation in DTA prediction. This, in turn, significantly impacts the models' generalization capabilities in the complex realm of drug discovery.

Results: To tackle these challenges, we introduce GLCN-DTA, a model specifically designed for proficiency in DTA tasks. GLCN-DTA innovatively integrates a graph learning module into the existing graph architecture. This module is designed to learn a soft adjacent matrix, which effectively and efficiently refines the contextual structure of protein and drug molecular graphs. This advancement allows for learning richer structural information from protein and drug molecular graphs via graph convolution, specifically tailored for DTA tasks, compared to the conventional fixed adjacent matrix approach. A series of experiments have been conducted to validate the efficacy of the proposed GLCN-DTA method across diverse scenarios. The results demonstrate that GLCN-DTA possesses advantages in terms of robustness and high accuracy.

Conclusions: The proposed GLCN-DTA model enhances DTA prediction performance by introducing a novel framework that synergizes graph learning operations with graph convolution operations, thereby achieving richer representations. GLCN-DTA does not distinguish between different protein classifications, including structurally ordered and intrinsically disordered proteins, focusing instead on improving feature representation. Therefore, its applicability scope may be more effective in scenarios involving structurally ordered proteins, while potentially being limited in contexts with intrinsically disordered proteins.

Keywords: Deep learning; Drug discovery; Drug–target affinity prediction; Graph learning-convolutional networks.

Fig. 1

Summary of representative methods relevant to DTI and DTA prediction. Our methods apply to the scope of graph-based deep learning methods for DTA prediction

Fig. 2

The architecture of GLCN-DTA. For protein sequences, protein graphs are built upon contact maps derived from their sequences. In the case of molecules, their SMILES representations are utilized as the foundation for graph construction. Once these two graph representations are established, they are processed through two separate graph learning-convolutional networks (GLCN) to extract their respective graph-level features. These representations are then concatenated to predict the affinity via fully connected layers. ${\mathbf{V}}^l$ donates node embedding in the l-th graph convolution layer. ${\mathbf{H}}^l$ represents hidden features in the l-th graph convolution layer. $\mathbf{A}$ is soft adjacent matrix. $\oplus$ denotes concatenated operation

Fig. 3

The procedure of constructing a drug molecule graph. The SMILES representations are utilized for molecule graph construction. Due to the complexity of the drug graph, we have depicted only a portion of the full graph for clarity and ease of demonstration

Fig. 4

The procedure of constructing a protein graph. The protein sequence are utilized for protein graph construction

Fig. 5

Ablation study for various dropout probabilities on the Davis datasets with GLCN-DTA methods