The CAS Registry® file is the most authoritative collection of chemical substances in the world. It has been built over more than 60 years from paper structure sheets to basic online registration, stereoregistration, and specialist review and corrections. As any chemist knows, but developers may not appreciate, there can be multiple ways to draw the same chemical substance that are equally correct. Conversely, there can be subtle, but crucial differences between compounds that must be taken into account when determining if it is a new or existing registration. And as powerful as our existing registration software is, there are limitations that mean our customers do not always get the most accurate chemical representation possible. Collaborative efforts in Project Phoenix between Content Operations, Content Tech, and DAI employees in Columbus and across the globe are working to minimize these limitations and build the Registry of the next 60 years. At the core of this effort is a modern Apache Spark pipeline on AWS EMR that processes hundreds of millions of substance records at scale, driven by graph-theoretic algorithms that transform chemical structure representations between formats and compute structural identity through layered canonicalization correctly unifying substances drawn differently while distinguishing subtly distinct compounds.
Central to the new architecture is the Component Expression, a unified XML representation capturing each substance's complete structure, stereochemistry, and metadata in a single schema-validated document, which serves as both the standardized registry format and the direct input for holistic identity determination across all components of a substance. Equally important is the deliberate modernization of the underlying codebase, translating decades of chemical logic from legacy C and proprietary systems into maintainable, well-tested Java and Python, enabling registry rule changes to be implemented and deployed in days rather than months, and building a platform the next generation of CAS developers can confidently build upon.
