The Information Machine
Updated today·New·first covered 23 Sep 2026·2 sources

RetroChimera retrosynthesis AI

The gist

Microsoft Research, GSK, Novartis publish RetroChimera retrosynthesis AI

Training on proprietary pharmaceutical data alongside public datasets is an unusual collaboration between a technology company and major drug manufacturers. The model is aimed at accelerating chemical synthesis planning, which is described as slow and expensive.

The full picture

A joint team from Microsoft Research, GSK, and Novartis published RetroChimera, an AI model for retrosynthesis planning, in a Nature paper. The model was trained on a combination of public and proprietary pharmaceutical reaction datasets and improved retrosynthesis predictions used in drug synthesis planning. Retrosynthesis, the process of working backward from a target molecule to find viable synthetic routes, is a key bottleneck in producing custom molecules for medicine, materials science, and agriculture.

How it developed
23 September 2026

Microsoft Research, GSK, and Novartis published RetroChimera in Nature on September 22, targeting the retrosynthesis planning step that determines how to synthesize a target molecule from existing reactions.

A distinctive feature is that the model was trained on both public and proprietary pharmaceutical reaction data, reflecting the collaboration between a technology research lab and two major pharmaceutical companies. RetroChimera improved retrosynthesis predictions and is designed to help researchers explore a wider range of molecules.

First citedMicrosoft Research TwitterThe Neuron
22 September 2026

The Neuron Daily reports RetroChimera as a joint Microsoft Research, GSK, and Novartis collaboration trained on public and proprietary data

21 September 2026

Microsoft Research posts about RetroChimera Nature paper on retrosynthesis AI

Sources
Microsoft Research Twitter
The Neuron
The daily email

Want this in your inbox?

I send one email each morning with the stories that moved. If you would rather just read here, that works too.

Subscribe free