Computational Landscape in Drug Discovery: From AI/ML Models to Translational Application

The combination of artificial intelligence (AI) and machine learning (ML) in drug discovery has significantly transformed traditional pharmaceutical research by enabling data-driven decision-making, accelerating the identification of hits, and improving the efficiency of lead optimization. This review provides a comprehensive overview of AI/ML models, including supervised, unsupervised, semisupervised, deep learning, and reinforcement learning approaches and their applications across various stages of drug development, from target identification and virtual screening to de novo molecule design and ADME/T prediction. We highlight widely used ML algorithms, performance evaluation metrics, and AI-driven tools that have become instrumental in modern drug discovery pipelines. Despite rapid advancements, challenges such as limited data availability, heterogeneity, bias, lack of model interpretability, reproducibility concerns, clinical translational barriers, and regulatory uncertainties continue to hinder full-scale adoption. The review also discusses emerging trends, including explainable AI, federated learning, and integration with high-throughput experimental platforms, which offer promising directions for overcoming current limitations. Emphasis is placed on the importance of interdisciplinary collaboration to bridge computational predictions with experimental validation, ensuring robust, ethical, and clinically translatable AI applications in drug development.