AMMVF-DTI: A Novel Model Predicting Drug-Target Interactions Based on Attention Mechanism and Multi-View Fusion

Abstract

Accurate identification of potential drug-target interactions (DTIs) is a crucial task in drug development and repositioning. Despite the remarkable progress achieved in recent years, improving the performance of DTI prediction still presents significant challenges. In this study, we propose a novel end-to-end deep learning model called AMMVF-DTI (attention mechanism and multi-view fusion), which leverages a multi-head self-attention mechanism to explore varying degrees of interaction between drugs and target proteins. More importantly, AMMVF-DTI extracts interactive features between drugs and proteins from both node-level and graph-level embeddings, enabling a more effective modeling of DTIs. This advantage is generally lacking in existing DTI prediction models. Consequently, when compared to many of the start-of-the-art methods, AMMVF-DTI demonstrated excellent performance on the human, C. elegans, and DrugBank baseline datasets, which can be attributed to its ability to incorporate interactive information and mine features from both local and global structures. The results from additional ablation experiments also confirmed the importance of each module in our AMMVF-DTI model. Finally, a case study is presented utilizing our model for COVID-19-related DTI prediction. We believe the AMMVF-DTI model can not only achieve reasonable accuracy in DTI prediction, but also provide insights into the understanding of potential interactions between drugs and targets.

Keywords: drug repositioning; drug–target interaction; graph attention networks; multi-head self-attention mechanism; neural tensor networks.

Figure 1

Results of ablation experiments on the human and C. elegans datasets in terms of AUC, precision, and recall.

Figure 2

Predicted scores of COVID-19-related DTIs using our AMMVF-DTI model. On the left, red triangles represent the interactions between the drug baricitinib and its 12 target proteins, while blue inverted triangles represent the interactions between the unrelated drug trazodone and the same proteins. On the right, red triangles represent the interactions between the drug remdesivir and its 16 target proteins, while blue inverted triangles represent the interactions between the unrelated drug aspirin and the same target proteins.

Figure 3

Distribution of (a) the bond numbers of drug molecules in the three datasets (left panel), (b) the molecular masses of drug molecules in the three datasets (middle panel), and (c) the partition coefficients of drug molecules (LogP) in the three datasets (right panel).

Figure 4

Framework of the proposed model AMMVF-DTI, which consists of three core modules: (1) feature extraction modules (BERT/GAT/ATT), (2) interaction information extraction modules (ITM/NTN), and (3) prediction module (MLP).

Figure 5

Structure of the BERT module, which consists of three parts: input embedding with multiple transformer encoding layers, a multi-head self-attention mechanism, and a feedforward neural network. First, the input vector X_n is transformed into vector Z_n through multi-head self-attention, and the two vectors are then added together using a residual connection. Subsequently, layer normalization and linear transformation are applied to the vectors to enhance the model’s capacity to capture complex patterns.

Figure 6

An overview of the structure of the ITM module. “Add” represents residue connection, “Norm” represents normalization, and “× N” represents the number of layers in the cross-attention mechanism.

Figure 7

An overview of the structure of the NTN module.