Pubchem 10M Canonicalized: Molecular Structures for Machine Learning

A processed version of the PubChem-10M dataset, canonicalized using RDKit and split into training and validation sets. The dataset was created by user 'sagawa' and last updated on September 4, 2022. It contains molecular structures represented as canonical SMILES strings.

Use Cases

• Train molecular property prediction models based on SMILES representations.
• Benchmark generative models for molecular design using the canonicalized structures.
• Validate cheminformatics algorithms on a large, processed dataset.
• Fine-tune language models on chemical structure data.

Strengths

• Data is canonicalized for consistency, removing SMILES unreadable by RDKit.
• Dataset is pre-split into training and validation subsets with a 9:1 ratio.

Limitations

• Description metadata is limited; actual data quality requires manual inspection after download.
• Column-level documentation is absent; field semantics must be inferred after download.
• Last updated 2022-09-04 02:18:37; freshness should be verified.

Provenance

Source

PubChem-10M dataset.

Collection Method

Downloaded and canonicalized using an RDKit-based function.

Freshness

Last updated 2022-09-04 02:18:37.