QSAR Bioconcentration Classes Dataset: 779 Chemicals with Mechanistic Labels

779 manually-curated chemicals were used to predict bioconcentration mechanisms, such as storage in lipid tissues or metabolism. The dataset includes Dragon descriptors, CAS numbers, SMILES strings, experimental BCF, and KOW values. It was developed by Grisoni et al. in 2016 for QSAR classification tree modeling.

Use Cases

• Predicting bioconcentration mechanisms (lipid storage, protein binding, metabolism) based on chemical descriptors.
• Training classification models like CART trees coupled with Genetic Algorithms.
• Comparing experimental and predicted octanol-water partition coefficients (KOW).
• Studying relationships between chemical structure (via SMILES) and environmental persistence.

Strengths

• Dataset contains 779 manually-curated chemicals with experimental bioconcentration factor (BCF) data.
• Includes 9 selected Dragon descriptors along with CAS and SMILES identifiers.
• Data was split into a training set of 584 compounds and a test set of 195 compounds, preserving class proportions.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Row count is known, but the number of features per row beyond the 9 descriptors is unclear.
• Last update date is unknown; freshness unverified.

Provenance

Source

UCI Machine Learning Repository, originally from Grisoni et al. (2016).

Collection Method

Manually-curated experimental data used for QSAR model development.