MOSAIC: Yale and Boehringer Ingelheim Launch AI Platform for Chemical Synthesis

Chemists at Yale, in collaboration with researchers from Boehringer Ingelheim Pharmaceuticals, have developed an AI-powered platform named MOSAIC. This framework is designed to generate experimental procedures for chemical synthesis, including for compounds that do not yet exist. The research was led by Yale's Victor Batista, co-corresponding author of a new study published in the journal Nature. Newhouse, a professor of chemistry at Yale and co-corresponding author, highlighted its role in simplifying synthetic chemistry.

This framework is designed to generate experimental procedures for chemical synthesis, including for compounds that do not yet exist.

A Novel Approach to Chemical Synthesis

MOSAIC aims to address the bottleneck in utilizing the vast amount of accumulated chemical reaction protocols. It is powered by 2,498 individual AI "experts," each representing knowledge in a specific chemistry-related topic. The developers state that this approach allows users to access expertise from thousands of distinct chemical reaction niches, outperforming existing AI chemistry systems that rely on a single, large model.

Demonstrating Real-World Capabilities

The Yale team demonstrated MOSAIC's capabilities by successfully synthesizing over 35 previously unreported compounds. The platform also provides users with measurable uncertainty estimates, indicating how well a request fits within an expert's domain of experience, which can help in prioritizing experiments.

The Future of AI in Experimental Chemistry

MOSAIC is fully open-source and compatible with future models. The researchers intend for it to facilitate the transition of AI from prediction to direct support for real-world experimentation in chemistry. Support for the study was provided, in part, by Boehringer-Ingelheim Pharmaceuticals and the National Science Foundation Engines Development Award.