The integration of artificial intelligence (AI) and big data is poised to significantly augment drug research and development, offering the potential to address persistent challenges such as lengthy timelines and high failure rates. This review provides a critical overview of AI applications across the preclinical drug discovery pipeline (the 2020–2026 literature), covering drug–target interaction prediction, structure prediction, de novo design, virtual screening, drug repurposing, and ADMET forecasting. Beyond surveying technical developments, we critically discuss key translational hurdles, including data quality, model interpretability, patient heterogeneity, and regulatory adaptation, and provide structured summaries of representative models. We conclude by outlining future directions, such as multimodal AI, digital twins, and closed-loop automation, that aim to bridge the gap between computational prediction and clinical application. This review aims to inform researchers and accelerate the delivery of safe and effective therapies.
Cai et al. (2026) studied this question.