R1: A Foundational Model for Small Molecule Drug Discovery (rhizomeresearch.ai)

Rhizome has announced the launch of r1, a foundational model for small molecule drug discovery, marking the first in a series designed to enhance drug design processes. This model addresses critical challenges in fragment-based generation by enabling precise customization, allowing researchers to implement native scaffold hopping, R-group exploration, and linker design effectively. With 245 million parameters, r1 is trained on approximately 800 million molecules, providing it with a deep understanding of chemical structures and allowing it to generate novel molecular species while adhering to user-defined constraints. The significance of r1 lies in its innovative approach to generating drug-like candidates, particularly in relation to existing medications targeting the EZH2 protein, a crucial player in cancer progression. By employing a core replacement strategy, r1 successfully identifies novel analogs with improved predicted binding affinities, outperforming traditional methods that lack the ability to control fragment characteristics effectively. This model not only expedites the search for effective drug candidates but also enhances the strategic maneuvering in intellectual property (IP) considerations, offering researchers a significant tool for navigating the complex landscape of drug development while optimizing patent strategies. The integration of advanced generative techniques with physics-based evaluations positions r1 as a game-changer in the realm of AI-assisted drug discovery.