SAVI Space

Taking the Step After Enumeration with the SAVI Space

The SAVI project was initiated at the U.S. National Institutes of Health (NIH), specifically the NCI CADD group, with the goal of creating a very large collection of synthetically accessible virtual molecules to support computational drug discovery and enable large-scale virtual screening.
Taking the original enumerated SAVI concept further, researchers at the ZBH (University of Hamburg) realized SAVI as a combinatorial Chemical Space built from Enamine building blocks, proving that ultra-large compound collections can be encoded efficiently using synthon-based representations.
Reaction-Driven Chemistry

Reaction-Driven Chemistry

To create SAVI Space, curated LHASA reaction transforms were applied to commercially available building blocks and translated into reaction SMARTS patterns. Instead of generating and storing every product molecule, reactions and compatible synthons are combined combinatorially during search, allowing billions of compounds to be represented in a compact format.

The SAVI Space version encodes more than 7.5 billion molecules while requiring only minimal storage and computational resources. Filtering rules derived from reaction knowledge ensure chemical stability and realistic synthetic feasibility.
Download
7.5 x 109
SAVI Space
Traits
  • Reaction-driven chemical space encoding.
  • High synthetic accessibility by design.
  • Efficient billion-scale exploration on standard hardware.
  • Complementary chemical diversity.
Additional Notes

What Makes SAVI Space Special

The development of the SAVI Space marks an important shift in how chemical libraries can be handled computationally: instead of generating and managing billions of individual molecules, the underlying reaction knowledge and compatible synthons are captured once and used to create compounds only when needed.

As a result, vast regions of synthetically accessible chemistry become searchable on standard hardware, enabling scalable similarity search, substructure exploration, and structure-based screening workflows. By proving that billion-scale, synthesis-aware libraries can be represented compactly without losing chemical realism, SAVI Space established a practical foundation for navigating ultra-large chemical spaces and significantly lowered the barrier to exploring new chemical matter in early drug discovery.

SAVI Versions — Enumerated and Combinatorial

Version Building blocks used Number of products Products without KILL Computational time* Storage size
SAVI-Lib-2020 (enumerated library) 1.55 × 105 > 1 × 109 — HPC (millions CPU h) ~210 GB
SAVI-Space-2020 (Lib-2020 rules) 1.39 × 105 2.34 × 109 3.65 × 109 3 h 2.1 GB
SAVI-Space-2020 1.38 × 105 2.40 × 109 3.34 × 109 3 h 0.8 GB
SAVI-Space-2024 2.56 × 105 7.55 × 109 1.07 × 1010 10 h 1.4 GB
The table highlights the transition from the explicitly enumerated SAVI-Lib-2020 to the combinatorial SAVI Space representations. While the original library required the generation and storage of more than a billion individual molecules, demanding high-performance computing resources and large storage capacity, the SAVI Space encodes the same underlying chemistry through reaction rules and compatible synthons.

This synthon-based representation drastically reduces storage requirements while enabling the exploration of even larger numbers of synthetically accessible compounds.
This highlights a fundamental shift in how ultra-large molecular collections can be used in practice: instead of handling static compound databases, researchers can dynamically generate and screen chemically realistic molecules, making large-scale structure-based discovery workflows significantly more efficient and accessible.

Liking a Compound from the SAVI Space?
Order It!

The space was constructed using more than 250,000 commercially available Enamine building blocks combined through 53 curated reactions, ensuring that generated molecules follow realistic and well-established synthetic routes.
Compounds discovered in the SAVI Space are not purely theoretical designs but are rooted in practical synthetic chemistry.
Because SAVI Space compounds are designed to be highly synthetically accessible and are directly derived from Enamine reagents, it is very likely that many identified hits can be synthesized on demand. In practice, promising candidates can often be produced and shipped by Enamine upon request, providing a direct path from virtual discovery to experimentally testable molecules.
Access with infiniSee

Access with infiniSee

The SAVI Space can be conveniently explored using BioSolveIT's Chemical Space navigation platform, infiniSee. Billions of entries can be searched for relevant chemistry within seconds. Various search algorithms are available to identify the best matches for the objectives of a drug discovery project.
Check out infiniSee
Want to Take It Into the 3D World?

Want to Take It Into the 3D World?

Chemical Space Docking® also enables structure-based screening of the SAVI Space within your target of interest. Executed within our drug design dashboard SeeSAR, molecules forming the most favorable interactions in the binding site can be efficiently mined on standard hardware, and without a connection to the internet.

This allows in-house projects to be streamlined without the involvement of third parties; you will work in a user-friendly environment with full control over your results.
Read about C-S-D

Eager to learn more?

More details on SAVI Space