AI reduces the cost of screening billions of molecular drugs to one thousandth

Researchers from institutions such as St. Jude Children’s Research Hospital in the United States have developed an artificial intelligence-assisted platform called AdaptiveFlow, aimed at significantly reducing the costs of drug development. This platform utilizes AI and efficient cloud computing architecture to reduce the virtual screening cost of compound libraries containing 69 billion molecules to one-thousandth of the original level. It also reduces the required computing resources to one-tenth of existing methods and supports linear expansion up to 5.6 million virtual CPUs. Its workflow includes pre-screening molecular grids by properties, identifying potential candidate molecules using machine learning models, and evaluating binding capabilities using molecular docking methods. Tests have shown that this platform can support over 1,500 molecular docking scenarios. In the validation of inhibitors targeting cancer-related targets, polyadenylic acid ribose polymerase 1 and ferroptosis inhibitor protein 1, the selected inhibitors met the standards for drug development. Currently, AdaptiveFlow has been made open-source and provides usage tutorials, aiming to provide new tools for low-cost super-large-scale virtual screening and research on difficult-to-develop drug targets.