Augmented Canonical Druglike QED 43M: Molecular Structures for Drug Discovery

A dataset of 43 million canonicalized molecular structures derived from the Druglike molecule datasets. The dataset was created by Derify and last updated on September 9, 2025. To enhance diversity, 33% of the dataset was randomly sampled and augmented using RDKit's Chem.MolToRandomSmilesVect function.

Use Cases

• Training molecular property prediction models based on canonical SMILES representations.
• Benchmarking molecular generation algorithms using the augmented SMILES subset.
• Developing and validating drug-likeness filters based on the QED (Quantitative Estimate of Drug-likeness) source data.

Strengths

• Contains 43 million molecular structures, providing a large-scale resource.
• 33% of the data was augmented to enhance molecular diversity.
• Structures were canonicalized using RDKit (version 2024.9.4) for consistency.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Row count is unknown, which may limit suitability assessment.
• Description metadata is limited; actual data quality requires manual inspection after download.

Provenance

Source

Derived from the Druglike molecule datasets for drug discovery.

Collection Method

Canonicalized using RDKit and augmented using RDKit's Chem.MolToRandomSmilesVect function, following an approach similar to NVIDIA's molmim method.

Freshness

Last updated 2025-09-09 22:27:37; freshness should be verified.