Скачать или смотреть Machine learning across the preclinical drug discovery pipeline

This video summarizes the Perspective “Machine Learning in Preclinical Drug Discovery” by Denise B. Catacutan, Jeremie Alexander, Autumn Arnold, and Jonathan M. Stokes (Nature Chemical Biology, 2024). The paper outlines how artificial intelligence is being integrated across the three main stages of preclinical drug discovery-hit identification, mechanism-of-action elucidation, and translational optimization. It reviews how deep learning and graph neural networks are accelerating virtual screening, how models such as Geneformer predict targets and mechanisms from single-cell transcriptomics, and how machine learning enhances ADMET prediction for solubility and toxicity during lead optimization. The authors emphasize that fully ML-integrated pipelines could reduce time and cost in early drug discovery while improving success rates. The Perspective also highlights the importance of model interpretability, open-source datasets, and collaboration between data scientists and experimental biologists for building reliable AI-driven discovery frameworks.