Pubchem_20260224_Filtered: 6 Million Compounds with Ro3 Filter and Descriptors

6,004,131 chemical compounds from a February 2026 PubChem snapshot have been filtered using the Rule of Three (Ro3) for drug-like properties. Each entry includes a PubChem ID, canonical SMILES string, and 43 molecular descriptors calculated with the RDKit cheminformatics toolkit. The dataset represents a curated subset of the original 123 million PubChem compounds, focused on lead-like chemical space.

Use Cases

• Virtual screening of drug-like compounds based on Rule of Three (Ro3) physicochemical properties.
• Training quantitative structure-activity relationship (QSAR) models using the 43 RDKit molecular descriptors.
• Linking chemical structures via PubChem IDs to external bioactivity databases for integrated analysis.
• Benchmarking cheminformatics pipelines and descriptor calculation methods against a large-scale standardized set.

Strengths

• Large scale with 6,004,131 pre-filtered compounds, enabling high-throughput computational studies.
• Explicit application of the Rule of Three (Ro3) filter (MW 50-300, CLogP<3, etc.) ensures focus on lead-like molecules.
• Includes 43 calculated molecular descriptors, adding immediate analytical value beyond basic identifiers.
• Cross-platform presence on Figshare suggests community validation and importance.

Limitations

• Column names and specific descriptor list are not provided in the available metadata.
• The dataset is a static snapshot from February 2026 and may not reflect later updates to PubChem.
• Lacks associated experimental data (e.g., bioactivity), limiting use to structural and property-based analysis.

Provenance

Source

PubChem (snapshot dated 2026-02 24)

Collection Method

Filtered from original PubChem snapshot using Rule of Three criteria; descriptors calculated with RDKit based on SMILES.

Time Range

Snapshot from 2026-02-24

Freshness

2026-03-22

Geography

null