New Paper: SAVI Space

22. Januar 2026

Foto: Malte Korn

The SAVI library is a collection of over a billion synthetically accessible compounds developed at the NIH as an open alternative to commercial chemical spaces. However, handling this vast collection explicitly is computationally demanding for drug discovery applications. To address this, we created SAVI-Space, a reaction-driven combinatorial chemical space, that encodes the transformation rules as reaction SMARTS. This approach yields 7.5 billion molecules while requiring only 1.4 GB of memory compared to 210 GB for the enumerated SAVI library. The improved search capabilities, including fast similarity and substructure searches on standard hardware. This represents a significant advance over the original library. Details on the design and implementation are now published in Nature Scientific Data.

The dataset and code are available at https://doi.org/10.25592/uhhfdm.15990 and https://github.com/rareylab/SAVI-Space.

The publication can be found here: https://www.nature.com/articles/s41597-025-05384-z / https://doi.org/10.1038/s41597-025-05384-z