Scientists Teach AI To Think Like a Professional Chemist
Researchers have developed a framework called Synthegy that uses large language models to guide computational methods in chemistry, enabling AI-assisted discovery. The system interprets chemical strategies written in natural language and evaluates reaction routes based on user goals.
Designing molecules is a difficult task in chemistry, requiring deep technical knowledge and strategic thinking. Researchers led by Philippe Schwaller at EPFL have introduced a framework called Synthegy that combines traditional search algorithms with artificial intelligence to interpret chemical strategies written in natural language. Synthegy starts with a target molecule and a plain language instruction from the user, generating possible reaction routes that are reviewed by a language model. The system scores each pathway based on how well it matches the user's goals and explains its reasoning. In synthesis planning tests, Synthegy identified routes that matched complex strategic requirements, with expert assessments agreeing with the system's results 71.2 percent of the time on average.
This content was automatically generated and/or translated by AI. It may contain inaccuracies. Please refer to the original sources for verification.