AI-driven drug–target interaction prediction: current progress, challenges, and future roadmap for precision medicine
- 31 March 2026
- journal article
- review article
- Published by Springer Nature in Journal of Computer-Aided Molecular Design
- Vol. 40 (1) , 89
Abstract
Drug–target interactions (DTIs) are fundamental to drug discovery, development, and repositioning. However, experimental methods for DTI identification are often constrained by high costs, time demands, and scalability issues, prompting a shift toward computational approaches. This review systematically explores recent advancements in computational DTI prediction, encompassing ligand-based, target-based, network-based, machine learning (ML), deep learning (DL), and hybrid multi-omics models. Ligand-based techniques, such as QSAR and pharmacophore modeling, offer structure–activity insights but require known ligands. Target-based methods rely on molecular docking and binding site prediction, yet often suffer from incomplete or unknown protein structures. Network-based strategies utilize bipartite and heterogeneous graphs integrated with protein–protein interaction (PPI) networks to infer novel DTIs. ML and DL methods especially graph neural networks and Transformer-based models have significantly improved prediction accuracy by leveraging chemical, biological, and omics features. Notably, hybrid models that integrate genomics, transcriptomics, proteomics, and interactomics data offer a systems biology perspective, enabling context-specific and personalized predictions. Key databases, like DrugBank, ChEMBL, and Binding DB, and computational tools such as Deep Purpose, NeoDTI, and FusionDTI, exemplify the latest advances in DTI prediction. Validation strategies are discussed through contemporary case studies. While substantial progress has been made, key challenges remain, including data sparsity, model interpretability, and generalization. Looking forward, emerging trends such as federated learning, AlphaFold-based docking, and quantum simulations are poised to further transform the field. This review emphasizes the importance of interdisciplinary integration and ethical frameworks, charting a roadmap for future DTI research and its translational applications in precision medicine. Graphical abstract Overview of machine-learning based approaches for predicting drug–target interactions using integrated drug, target, and interaction databases, highlighting supervised and semi-supervised learning frameworks and their key methodological subclassesKeywords
This publication has 298 references indexed in Scilit:
- Controllable Edge-Type-Specific Interpretation in Multi-relational Graph Neural Networks for Drug Response Prediction2025
- Identification of Acetylcholinesterase Inhibitors Through a Pharmacophore‐Guided Deep Learning Approach for Therapeutic Applications in Alzheimer's DiseaseChemistrySelect, 2025
- MFAE: Multilevel Feature Aggregation Enhanced Drug‐Target Affinity Prediction for Drug Repurposing Against Colorectal CancerAdvanced Intelligent Systems, 2023
- Applications of multi‐omics analysis in human diseasesMedcomm, 2023
- Role of protein phosphorylation in cell signaling, disease, and the intervention therapyMedcomm, 2022
- Target Identification Approaches in Drug Discovery2022
- The Structure and Mechanism of Drug Transporters2021
- Structure based virtual screening: Fast and slowWIREs Computational Molecular Science, 2021
- Current advances in ligand‐based target predictionWIREs Computational Molecular Science, 2020
- QSAR Modeling of SARS‐CoV Mpro Inhibitors Identifies Sufugolix, Cenicriviroc, Proglumetacin, and other Drugs as Candidates for Repurposing against SARS‐CoV‐2Molecular Informatics, 2020