USearch Molecules: 7 Billion Small Molecules with Pre-Built Search Indexes

7 billion small molecules are represented in SMILES notation, accompanied by 28 billion molecular fingerprints including MACCS, ECFP4, FCFP4, and PubChem types. The collection includes pre-constructed USearch indexes for efficient similarity search. It is hosted on AWS Open Data and published by Ash Vardanian under an Apache-2.0 license.

Use Cases

• Perform molecular similarity searches based on pre-built USearch indexes over fingerprints.
• Conduct virtual screening of chemical libraries based on the 7 billion SMILES representations.
• Benchmark fingerprinting algorithms based on the inclusion of multiple fingerprint types (MACCS, ECFP4, FCFP4, PubChem).
• Train machine learning models for molecular property prediction based on the large-scale structural data.

Strengths

• Contains 7 billion small molecules, providing an extremely large chemical space.
• Includes 28 billion pre-computed molecular fingerprints across four established types.
• Offers pre-built USearch indexes, which can accelerate similarity search operations.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Row count for the fingerprint data is unknown, which may limit suitability assessment.
• Last update date is unknown; freshness unverified.

Provenance

Source

Ash Vardanian via AWS Open Data.

Collection Method

Collection and indexing of molecular structures and fingerprints; specific methodology is not detailed.

Time Range

Temporal coverage is unknown.

Freshness

Last update date is unknown; freshness unverified.

Geography

Spatial coverage is unknown.