BindingSiteDTI: differential-scale binding site modelling for drug-target interaction prediction

Abstract

Motivation: Enhanced by contemporary computational advances, the prediction of drug-target interactions (DTIs) has become crucial in developing de novo and effective drugs. Existing deep learning approaches to DTI prediction are frequently beleaguered by a tendency to overfit specific molecular representations, which significantly impedes their predictive reliability and utility in novel drug discovery contexts. Furthermore, existing DTI networks often disregard the molecular size variance between macro molecules (targets) and micro molecules (drugs) by treating them at an equivalent scale that undermines the accurate elucidation of their interaction.

Results: We propose a novel DTI network with a differential-scale scheme to model the binding site for enhancing DTI prediction, which is named as BindingSiteDTI. It explicitly extracts multiscale substructures from targets with different scales of molecular size and fixed-scale substructures from drugs, facilitating the identification of structurally similar substructural tokens, and models the concealed relationships at the substructural level to construct interaction feature. Experiments conducted on popular benchmarks, including DUD-E, human, and BindingDB, shown that BindingSiteDTI contains significant improvements compared with recent DTI prediction methods.

Availability and implementation: The source code of BindingSiteDTI can be accessed at https://github.com/MagicPF/BindingSiteDTI.

Figure 1.

A protein interacts with different drugs, binding sites are extremely different due to the size of drugs.

Figure 2.

BindingSiteDTI includes two main modules: MMCMSE and MMSID. MMCMSE extracts fixed-scale substructures (block a) for the drug and multiscale substructure for the target (block c). Top K% relevant substructure between drugs and targets will be selected as the preliminary step of interaction modelling (block b). MMSID will learn the hidden relationship between different drugs and target substructures as the interaction feature for downstream prediction.

Figure 3.

Protein tokens visualization: we trace protein tokens back to original nodes and visualized by UCSF Chimera (Pettersen et al. 2004)

Figure 4.

Experiment results on human dataset. Average AUROC and AUPRC are illustrated with error bars of standard deviation.

Figure 5.

The visualization result of BindingSiteDTI during prediction unseen drug–target pairs, the highlighted red region is top k attention weighted substructure of target by the proposed method. For each case, we plot the ground truth of the binding site on the left and our visualization on the right.