System and method for query-based random access to a virtual chemical combinatorial synthesis library
Patent Information
- Authority / Receiving Office
- JP · JP
- Patent Type
- Applications
- Current Assignee / Owner
- ATOMWISE INC
- Filing Date
- 2023-05-16
- Publication Date
- 2026-05-25
AI Technical Summary
Existing methods for navigating and querying combinatorial synthesis libraries face challenges with scalability, chemical validity, and synthetic accessibility, particularly in non-enumerable chemical spaces, as they rely on exhaustive enumeration and struggle with long autoregressive chains and computational inefficiencies.
A system and method using a graph-based generative model that encodes molecular queries, allowing for query-based random access to large combinatorial synthesis libraries by learning a hierarchy of keys across the library components, minimizing autoregression, and enabling efficient parallelization, thus overcoming scalability and computational complexity issues.
The system provides efficient navigation and random access to vast non-enumerated compound libraries, ensuring chemical validity and synthetic accessibility with reduced computational complexity and parameter count, making it suitable for large molecular graphs.
Smart Images

Figure 00000000_0000_ABST