Saturn‑RXN: Generative Molecular Design with Steerable and Granular Synthesizability Control for Greener Chemical Synthesis
2026 Data Science and Modeling for Green Chemistry Award
The team at École Polytechnique Fédérale de Lausanne (EPFL) in Switzerland recognized for SaturnRXN, an advanced generative molecular design framework that brings sustainability constraints directly into AI driven discovery. While traditional generative models create novel structures without regard for real world synthesis, SaturnRXN integrates explicit reaction rules and green chemistry criteria into the design process. Using a sample efficient language model combined with reinforcement learning, the system proposes property optimized molecules that also satisfy key synthesis requirements such as avoiding hazardous transformations, minimizing steps, or mandating biobased or waste derived building blocks.
The approach directly aligns with core Green Chemistry principles, supporting atom economy, waste reduction, the use of renewable feedstocks, and safer reaction pathways. In a case study, SaturnRXN successfully designed viable new molecules derived from industrial waste streams, demonstrating its potential to create sustainable yet high value chemical matter.
This “sustainable by design” paradigm offers a transformative path for pharmaceutical and agrochemical innovation by coupling human expertise with AI guided, greener synthesis planning.
Team Members
-
Víctor Sabanza-Gil
-
Jeff Guo
More about the Award
The Data Science and Modeling for Green Chemistry Award aims to recognize the research and development of computational tools that guide the design of sustainable chemical processes and the execution of green chemistry that demonstrates compelling environmental, safety, and efficiency improvements over current technologies in the pharmaceutical industry and its allied industrial partners.
This annual award is presented at the Green Chemistry & Engineering Conference, where presenters are invited to share their innovations.
