IIC-DTI: A Contrastive Learning Enhanced Inter-Intra Molecular Fusing Framework for Drug-Target Interaction...

Area of Science:

• Bioinformatics
• Computational Chemistry
• Drug Discovery

Background:

• Accurate prediction of drug-target interactions (DTIs) is crucial for drug development and repositioning.
• Current DTI prediction models often overlook the complex associations between drugs and targets by analyzing their features independently.

Purpose of the Study:

• To develop an accurate and efficient DTI prediction method by integrating intramolecular and intermolecular features of drugs and targets.
• To address the limitations of existing models that ignore hidden associations between drug-target pairs.

Main Methods:

• A contrastive learning model, IIC-DTI, was designed to fuse intramolecular (drug chemical structures, target amino acid sequences) and intermolecular (drug-target pairs) features.
• A multi-head cross-attention network extracted intermolecular features, while a contrastive learning module fused information between different views of drug and target embeddings.
• The integrated embeddings were fed into a neural network for DTI prediction.

Main Results:

• IIC-DTI demonstrated superior performance compared to nine state-of-the-art methods, including large language models, across four benchmark datasets.
• A case study successfully validated 16 out of 20 predicted drug-target pairs through literature evidence.
• The model effectively identified relevant interactions for given drugs and targets.

Conclusions:

• IIC-DTI offers a promising approach for enhancing DTI prediction accuracy by effectively leveraging both intramolecular and intermolecular information.
• The model's performance and validation suggest its potential applicability in realistic drug discovery and repositioning scenarios.