EcoTox-QSAR Predictor

Estimate aquatic or soil toxicity for chemical–species–endpoint–condition combinations without a matched measurement.

Separately fitted aquatic (W00) and soil (M00) models combine molecular structure, species, and available experimental context while retaining task-specific response definitions.

Read each estimate alongside available task performance and input-side reference support to help prioritize follow-up measurements.

Extrapolation requires caution

S1/S2 primarily evaluate conditional completion in the retained population; S3/S4 probe stronger structural separation. Treat low-support queries cautiously. Site flags are not calibrated estimates of individual prediction error.

Aquatic and soil predictions Descriptive support distance Single and batch prediction

Single Prediction

Compound and exposure

SMILES Use canonical SMILES when possible; chemical support is assessed for the organic parent structure. Exposure medium Exposure time h Keep the exposure duration consistent with the endpoint record when known.

Prediction task

Prediction model Endpoint Effect EC level

EC predictions default to EC50. The selected medium limits the available model and local species list.

Species and taxonomy

Species Latin name search

Local species dropdown Matched species Kingdom Phylum Class Order Family Genus Species Full Latin name

Batch Prediction

Download template

Batch files are parsed only for this prediction request; this MVP does not retain uploads for incremental training.

SMILES	Task	Model-scale mean	Support note	Support tier

Prediction Result

Run a prediction to show results.

pTox toxicity index -

mol/L (aquatic concentration) -

mg/L (MW-converted concentration) -

The model-scale value is the transformed target. Aquatic predictions use -log10(mol/L); soil predictions use -log10(mol/kg) and are converted to mg/kg for display.

Chemical distance -

Species distance -

Task distance -

Overall AD distance -

This distance compares the input with the site's local reference data. It is descriptive, not a calibrated error probability or a pass/fail threshold; support tiers are heuristic.

Subtask R2 -

Subtask RMSE -

Subtask MAE -

Evaluation n -

Metrics summarize available evaluation records for the selected task, not uncertainty for this individual prediction. The evaluation subset and sample count are shown below; unavailable values remain blank.

Subtask Prediction Performance

Observed vs predicted model-scale value

The star shows this input's predicted value on the predicted axis only. No observed value is known for this input; the plot is historical evaluation data, not evidence that this input has been validated.

Task head: -
Metric source: -
AD warning: -
Training ring-scaffold status: -

Anonymous usage paths and service timings are retained for up to 30 days. Chemical inputs and files are excluded from analytics; queued inputs and results expire after one hour.