Russian AI Model to Accelerate Drug Discovery by Dozens of Times
The Matcha system enables fast and accurate molecular modeling, significantly speeding up the search for new medicines.
Researchers at Skoltech have developed a high-precision AI-based solution for molecular modeling that can speed up drug discovery by up to 30 times, the university said in a statement. The technology opens up new opportunities for global medicine.
The researchers explained that diseases occur when one or more proteins stop functioning properly, making them therapeutic targets for drugs. Drug molecules bind to these proteins with high specificity. Computational modeling allows scientists to test how well a molecule fits a particular protein in terms of shape and chemical properties. Docking technology has become a core part of virtual screening, where millions of compounds are evaluated to identify promising drug candidates.
Fitting Molecules to Targets
According to the team, candidate molecules undergo virtual testing before entering the laboratory. The Matcha model iteratively refines the positioning of a molecule within a protein pocket. Unrealistic configurations are discarded early, while a separate neural network selects the most optimal arrangement.
In parallel, the researchers are developing tools for other tasks, including molecule generation, prediction and optimization of their properties.
