Retrosynformer: planning multi-step chemical synthesis routes via a decision transformer

2025/11/18 by Emma Granqvist, Rocío Mercado, Samuel Genheden · 1 voice
Materials Science · Environmental Science · Engineering · #Machine Learning in Materials Science #Chemistry and Chemical Engineering #Process Optimization and Integration

paper · pdf · doi:10.1039/d5dd00153f

Abstract

We present RetroSynFormer, a novel approach to multi-step retrosynthesis planning. Here, we express the task of iteratively breaking down a compound into building blocks as a sequence-modeling problem and train a Decision Transformer-based model.

Citations

• A simple similarity metric for comparing synthetic routes

Cited by

• TempRe: Template generation for single and direct multi-step retrosynthesis

Discussions

• Lots of exciting news this week! First, our PhD student Emma's RetroSynFormer work on multi-step retrosynthesis planning with a Decision Transformer was accepted in @digital-discovery.rsc.org! 🥳 #che [bsky, 7 points, 1 comments]

Related

• RetroSynFormer: Planning multi-step chemical synthesis routes via a Decision Transformer
• Predicting retrosynthetic pathways using a combined linguistic model and hyper-graph exploration strategy
• Towards "AlphaChem": Chemical Synthesis Planning with Tree Search and Deep Neural Network Policies
• ChemiRise: a data-driven retrosynthesis engine
• RetroXpert: Decompose Retrosynthesis Prediction like a Chemist