AI lets chemists design molecules by simply describing them
AI Summary: Researchers at EPFL, led by Philippe Schwaller, have developed Synthegy, a novel framework that utilizes large language models (LLMs) to enhance retrosynthesis and reaction mechanism planning in chemistry. Instead of generating chemical structures, Synthegy evaluates potential synthetic pathways based on natural language instructions from chemists, allowing for more intuitive and efficient exploration of complex chemical strategies. In a double-blind study involving 36 chemists, Synthegy's evaluations aligned with expert assessments 71.2% of the time, demonstrating its effectiveness in identifying feasible pathways and improving the decision-making process in chemical synthesis. This approach signifies a shift in how AI can assist chemists by integrating strategic reasoning with computational tools.