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

Area of Science:

• Computational chemistry
• Bioinformatics
• Drug discovery

Background:

• Accurate drug-target interaction (DTI) prediction is vital for novel drug development.
• Current deep learning methods for DTI prediction often overfit molecular representations and fail to account for size differences between drugs and targets.
• This limitation hinders predictive reliability and practical application in discovering new therapeutics.

Purpose of the Study:

• To propose a novel deep learning network, BindingSiteDTI, designed to enhance DTI prediction.
• To address the limitations of existing methods by incorporating a differential-scale scheme to model the binding site.
• To improve the accuracy and reliability of DTI prediction for drug discovery.

Main Methods:

• Developed BindingSiteDTI, a novel deep learning network incorporating a differential-scale scheme.
• Extracted multiscale substructures from targets and fixed-scale substructures from drugs to model their interactions.
• Facilitated the identification of structurally similar substructural tokens and modeled relationships at the substructural level.

Main Results:

• BindingSiteDTI demonstrated significant improvements in DTI prediction accuracy compared to existing methods.
• Experiments were conducted on popular benchmarks including DUD-E, human, and BindingDB.
• The proposed differential-scale scheme effectively enhances the modeling of drug-target binding sites.

Conclusions:

• BindingSiteDTI offers a more reliable and accurate approach to drug-target interaction prediction.
• The differential-scale scheme is effective in handling molecular size variations between drugs and targets.
• This advancement holds promise for improving the efficiency and success rate of novel drug discovery.