AI-driven drug-target interaction prediction: current progress, challenges, and future roadmap for precision medicine

Area of Science:

• Pharmacology and Bioinformatics
• Computational Drug Discovery
• Systems Biology

Background:

• Experimental identification of drug-target interactions (DTIs) is costly, time-consuming, and difficult to scale.
• Computational approaches are increasingly vital for efficient DTI prediction in drug discovery and repositioning.
• Existing computational methods face challenges such as data sparsity and incomplete structural information.

Purpose of the Study:

• To systematically review recent advancements in computational drug-target interaction (DTI) prediction.
• To explore diverse methodologies including ligand-based, target-based, network-based, machine learning, deep learning, and hybrid multi-omics models.
• To highlight key databases, tools, challenges, and future trends in the field.

Main Methods:

• Review of ligand-based (QSAR, pharmacophore) and target-based (molecular docking) techniques.
• Analysis of network-based strategies integrating protein-protein interaction (PPI) networks.
• Evaluation of machine learning (ML) and deep learning (DL) models, including graph neural networks and Transformer-based approaches.
• Exploration of hybrid multi-omics models integrating diverse biological data.
• Discussion of validation strategies and contemporary case studies.

Main Results:

• Machine learning and deep learning models, particularly graph neural networks and Transformers, significantly enhance DTI prediction accuracy.
• Hybrid multi-omics models provide a systems biology perspective, enabling context-specific and personalized predictions.
• Key databases (DrugBank, ChEMBL) and tools (Deep Purpose, NeoDTI) represent the forefront of DTI prediction advancements.
• Despite progress, challenges in data sparsity, model interpretability, and generalization persist.

Conclusions:

• Computational DTI prediction is rapidly evolving, driven by ML, DL, and multi-omics integration.
• Future directions include federated learning, AlphaFold-based docking, and quantum simulations for further transformation.
• Interdisciplinary integration and ethical frameworks are crucial for advancing translational applications in precision medicine.