Business

Novartis and CAS build reaction-data backbone for AI drug discovery

Novartis and CAS build reaction-data backbone for AI drug discovery

CAS, a division of the American Chemical Society, and Novartis Biomedical Research announced a collaboration to build an AI-ready reaction-data resource for Novartis researchers.

The initiative will combine Novartis' proprietary experimental reaction records with more than 160 million scientist-curated reactions from the CAS Content Collection. CAS will use its Intelligence Hub to standardize and harmonize reaction data currently dispersed across electronic lab notebooks, reports, shared drives, and partner platforms, making the information more accessible for computational research, machine learning, and AI applications.

CAS will also work with Novartis to develop a customized discovery platform based on CAS SciFinder architecture, allowing researchers to search and analyze internal and external reaction data together. The collaboration is focused less on developing a new AI model than on creating the standardized chemistry-data layer needed to support AI-enabled discovery workflows.

CAS said the integrated resource could also provide a foundation for future large language model and agentic AI capabilities. The collaboration extends an existing relationship between CAS and Novartis and forms part of Novartis' broader effort to use AI and data science across drug discovery.

The AllSci BriefFree, systematic R&D and deal news. Daily.

The CAS project adds a data-infrastructure layer to Novartis’ broader AI-enabled discovery strategy. In July 2026, Novartis partnered with Chai Discovery to use generative AI models for therapeutic antibody design, while an expanded collaboration with Isomorphic Labs covers AI-enabled small-molecule discovery against up to six targets. Novartis has also struck AI-focused discovery deals with Relation Therapeutics in atopic diseases and Orionis Biosciences in molecular glues.


Access the AllSci platform to explore the science behind the news.


Spot something wrong? Report an issue with this article