ZINC Chemical Compounds for Virtual Screening

The ZINC dataset is a curated collection of commercially available chemical compounds for virtual screening. It is used for molecular property prediction, specifically as a graph regression task aiming to predict constrained solubility, with performance measured by MAE.

Use Cases

• Train graph neural networks for molecular property prediction using the constrained solubility target.
• Benchmark graph regression models on the ZINC dataset's chemical compound structures.
• Perform virtual screening by analyzing the predicted solubility of commercially available compounds.

Strengths

• Curated specifically for virtual screening applications.
• Dataset size is categorized as between 100K and 1M entries.

Limitations

• Specific row count, column details, and file formats are unknown.
• The dataset's temporal coverage and geographic origin are not specified.

Provenance

Source

graphs-datasets on Hugging Face.

Collection Method

A curated collection of commercially available chemical compounds, as described by Wikipedia.

Freshness

Last updated on 2023-02-07.

Geography

Region tag indicates 'us', but specific spatial coverage is unknown.