Abstract
Efforts to reduce costs and improve efficiency in drug discovery have driven the rapid integration of computational technologies into the pharmaceutical field. In particular, the quantitative advances in computational resources and data availability in recent years have enabled transformative innovations based on deep learning, fueling widespread discussions about the potential of AI under the term “AI drug discovery.” AlphaFold2, which has revolutionized the accuracy of protein tertiary structure prediction, serves as a landmark example of AI technology in drug discovery. Its applications span diverse areas, including peptide and antibody design, significantly expediting the identification and optimization of drug candidate molecules. Furthermore, advancements in language model technologies, originally pioneered in natural language processing, have facilitated sophisticated information representation through chemical and protein language models. These models support highly accurate predictions across a range of modalities and outcomes, offering unprecedented utility in the drug discovery process. The introduction of AlphaFold3 has further advanced the precision of protein-drug complex structure prediction, unlocking new opportunities for molecular design and therapeutic innovation. This review highlights the latest advancements in molecular design powered by AI technologies and examines their contributions to the increasingly complex and diverse landscape of modern drug discovery, supported by illustrative examples.
