Improving synthesis prediction of small molecules at scale with RetroChimera
Custom-made molecules are advancing medicine, materials, and agriculture, but producing them is slow and expensive. A new Nature paper highlights RetroChimera, a predictive model that helps accelerate chemical synthesis, helping researchers explore a wide range of molecules. The post Improving synthesis prediction of small molecules at scale with RetroChimera appeared first on Microsoft Research .
The research paper "Improving synthesis prediction of small molecules at scale with RetroChimera" describes a new framework for retrosynthesis prediction, published in the journal Nature. RetroChimera combines two complementary models to automatically propose high-quality synthesis routes, addressing the time-consuming and costly nature of chemical synthesis in the development of new medicines and materials.
The model uses R-SMILES 2, a Transformer-based de-novo model, and NeuralLoc, a graph neural network-based model, to generate predictions. R-SMILES 2 is better at reactions with large changes over the course of the reaction, while NeuralLoc tends to produce more accurate outputs for reactions not covered by a template library. The ensemble-based approach combines the strengths of both models, enabling RetroChimera to provide better predictions than either model alone.
Blind tests show that chemists prefer RetroChimera's individual reaction predictions over preceding models and literature reactions. This framework has the potential to accelerate the development of new medicinally relevant molecules and advanced materials by enabling efficient synthesis planning.
Written by urgent.news from Microsoft Research's reporting — not their text. Machine-written — may contain errors; check the original before relying on it.
