SpaceLight: A Spotlight on the Analog Hunter for Chemical Spaces

SpaceLight

Spotlight:

Hunting for close analogs in Chemical Spaces

SpaceLight: The novel method to retrieve accessible analogs from trillions of compounds

We are proud to feature SpaceLight — A novel Chemical Space exploration tool which screens vast combinatorial compound spaces for close analogs of a query compound.
SpaceLight was developed in collaboration with the University of Hamburg (ZBH) and utilizes topological, Tanimoto-type fingerprint similarity such as the well-known ECFPs.
After performing a search, retrieved similar compounds are scored, ranked, and provided as an enumerated list. SpaceLight is available as a command line tool and will become a component of infiniSee in the close future.

What are fingerprints?

Molecular fingerprints are essential cheminformatics tools to encode and describe structural features of compounds. They are used to screen for compounds similar to a query structure, and for other tasks such as clustering of compounds, diversity analyses, and detection of shared scaffolds.
Along the different fingerprint subtypes which are explained below, users can select several variants (also known as "iterations") of those to fine-tune their results. Variants describe the order of neighbor shells that encode the features of atoms in a molecule. The variant number indicates the shell size of the generated substructures used for the similarity search. Larger substructures cover more features and atom arrangements of the molecule, yet smaller differences receive a penalty in their similarity score which can affect the ranking of the results. Therefore, it can sometimes be required to fine-tune the screening by applying different variants of the fingerprint subtype.
For instance, the variants of ECFPs describe the effective diameter of the largest feature of iterations performed. An example given above shows which area of the structure corresponds to its ECFP variant.

What fingerprint subtypes can be searched with SpaceLight?

Several molecular topological fingerprint subtypes can be used in SpaceLight. Alongside the well-known ECFP4 fingerprint, new "CSFP" methods are available to retrieve compounds of interest from vast Chemical Spaces based on the challenges and scenarios related to individual drug discovery projects. ECFP considers only circular features for similarity.

ECFP: The extended-connectivity fingerprint. Its variant ECFP4 belongs to the best-performing fingerprints in small molecule virtual screening and target prediction benchmarks.^[1],[2],[3]

Further, SpaceLight also supports a variety of Connected Substructure (or Subgraph) Fingerprints (CSFPs) and their variants.

fCSFP: A fingerprint for fine-grained similarity measurement.
tCSFP: A topological fingerprint method with high potential for scaffold hopping.
iCSFP: An independent fingerprint method using the maximum-common substructure as the similarity descriptor.

Discover compounds for any scenario

Different fingerprint methods can be used for different needs of your project. Some examples of the diverse chemistry retrieved with SpaceLight and different fingerprints are displayed in the figure. Nemorexant, an FDA-approved orexin antagonist used for the treatment of insomnia, was used as the query compound.
The exploration of the Chemical Space around a query compound can deliver proposals for close analogs to investigate structure-activity relationships of the compound class, or alternatively, can be used to retrieve related but novel pharmacophores for scaffold hopping.

Source for accessible compounds

The molecules retrieved from dedicated Chemical Spaces with SpaceLight are accessible by synthesis. The molecules are not fantasy molecules that require complex synthesis, but instead they can be ordered from the compound makers — or be synthesized in one or two steps in-house. For this realized workflow, we collaborate with Enamine, WuXi LabNetwork, OTAVA, Chemspace and eMolecules in creating actionable make-on-demand Chemical Spaces.
The compounds can also be used in complementary enrichment methods: Use them for machine learning approaches or structure-based calculations (e.g., docking studies) to narrow down the best candidates for follow up.