Molecular Compound Database for Virtual Screening

SANQI is a collection of molecular databases and samples for computational chemistry and drug discovery, created by luskyqi. It includes multiple subsets such as ChemDiv, Enamine, FDA-approved compounds, and natural products, along with algorithmically generated samples. The dataset was last updated in March 2026.

Use Cases

• Train graph neural networks on molecular graphs from the 'bb_total.pkl' or 'np_total.pkl' files for property prediction tasks.
• Benchmark generative models using the 'qubo_sample' subsets (e.g., qubo_pick_NP.pkl) designed for specific chemical profiles.
• Perform virtual screening by comparing query molecules against the 'fda_total.pkl' database of approved drugs.
• Analyze chemical space coverage by comparing unique compound sets like 'ChemDiv_unique.pkl' and 'Enamine_unique.pkl'.

Strengths

• Contains multiple specialized subsets (e.g., FDA-approved, natural products, vendor libraries) for diverse benchmarking.
• Includes algorithmically curated samples (qubo_sample) for targeted chemical property exploration.

Limitations

• Unknown total number of molecular compounds or rows across all files.
• Data organization and file formats (e.g., .pkl) imply a dependency on specific Python libraries for access.

Provenance

Source

Compiled from vendor libraries (e.g., ChemDiv, Enamine), public drug databases, and algorithmic sampling.

Collection Method

Curated aggregation and algorithmic sampling (QUBO), stored in serialized pickle (.pkl) files.

Freshness

Last updated March 2026.