MarineTox Predictor: Ecotoxicity Data and Hazard Thresholds for 68,000 Chemicals

MarineTox Predictor is an online library platform containing 1.2 million records of predicted ecotoxicity data and derived environmental hazard thresholds. The dataset was developed by Yongle Zhu using multitask deep learning to predict 31 saltwater toxicity endpoints for 26 marine organisms, sharing knowledge from freshwater ecotoxicity data. It was last updated on 2026-04-21.

Use Cases

• Predicting saltwater ecotoxicity endpoints for chemicals based on shared substructure features from freshwater data.
• Identifying key chemical substructures with toxicity to specific marine organisms via the constructed interaction network.
• Deriving environmental hazard thresholds for chemicals to assess marine ecological risk.
• Screening chemicals of toxicity concern to marine ecosystems based on predicted data.

Strengths

• Contains 1.2 million records of ecotoxicity data and hazard thresholds.
• Covers predictions for 31 toxicity tasks across 26 marine organisms spanning five phyla.
• The model achieved an R² improvement of up to 140% for 18 low-resourced tasks compared to a baseline.
• Derived hazard thresholds for approximately 68,000 chemicals, identifying 902 as of toxicity concern.

Limitations

• Row count for the underlying dataset is unknown, which may limit suitability assessment.
• Column-level documentation is absent; field semantics must be inferred after download.
• The dataset is small in file size (251.7 KB), suggesting it may contain aggregated results or summary data rather than raw experimental records.

Provenance

Source

Yongle Zhu, via figshare.

Collection Method

Generated via multitask deep learning models with a cross-domain knowledge-sharing mechanism from freshwater ecotoxicity data.

Freshness

Last updated 2026-04-21 01:13:25; freshness should be verified.