ToxGPS - Toxicity Knowledgebase




ToxGPS® provides a reliable toxicity knowledgebase including workflows and predictions for a series of in vivo and in vitro human health and regulatory-relevant endpoints. The predictions are based on mechanistically-informed, probabilistic QSAR models and endpoint-specific structural knowledge (expert rules) which are combined to provide a weight-of-evidence prediction together with an estimate of the associated uncertainty. Nearest neighbors of the query compounds in the training sets are identified and linked to the experimental study data in the ChemTunes databases.

ToxGPS Predictions

ToxGPS® predictions are also suited for the assessment of genotoxic impurities in drug products under the ICH M7 guideline.

ToxGPS®also features the ToxGPS Read-Across workflow for data gap filling by identifying structural, physicochemical and biological analogs from the ChemTunes databases. The system offers tools to line up diverse evidence sources, either experimental or in silico, and to combine them to obtain an overall outcome with estimated uncertainty. Users can interactively explore how the read-across outcome varies depending on which evidence sources are used and the reliability assigned to each source.


[1] Cherkasov A, et al. QSAR modeling: where have you been? Where are you going to? J. Med. Chem. 2014, 57(12), 4977-5010.
[2] Leist M, Yang C, et al. Novel Technologies and an Overall Strategy to Allow Hazard Assessment and Risk Prediction of Chemicals, Cosmetics, and Drugs with Animal-Free Methods. Altex 2012, 29, 4/12, 373-388.
[3] Yang C, et al. Computational Toxicology Approaches at the US Food and Drug Administration. ATLA 2009, 37, 523-531.
[4] Rathman JF, et al. Uncertainty Estimation and Quantitative Combination-of-Evidence using Dempster-Shafer Theory. Comp. Tox. 2018, 6, 16-31.

Key Features

ToxGPS provides a reliable toxicity prediction system for a series of in vivo and in vitro human health and regulatory-relevant endpoints as well as workflows to support the safety and risk assessment of chemical compounds.

Available Prediction Endpoints

Available endpoints


  • QSAR outcomes by global and local mode-of-action (mechanistic) models
  • Chemotype alerts (structure knowledge) including severity scores (odds ratios)
  • Nearest neighbors of query compound including experimental study calls and link to ChemTunes Toxicity Database
  • Weight of evidence for combined final outcome based on decision theory approach minimizing and assigning uncertainties of predictions

Data Sources

  • US NTP, CPDB, US EPA ToxRefDB, US FDA, ILSI DevTox, EFSA, SCCS/SCCP/SCCNFP, ECETOC, HESS, Substance Registration database (ECHA) and scientific literature

Further Workflows

  • Read-across
  • TTC Evaluation
  • Metabolism prediction and prioritization (Liver BioPath)

User Interface

  • Web-based user interface through ChemTunes platform supporting standard web browsers


  • ToxGPS predictions

Prediction result with global and local models and structural (chemotype) alerts

screenshot ToxGPS Prediction

Prediction result with global and local models and nearest neighbors

screenshot ToxGPS Prediction

Read-Across assessment table

screenshot ToxGPS Read-Across


System Requirements

  • Linux (Ubuntu 14.04/16.04 LTS or RedHat 6/7)
  • Microsoft Windows 7, Server 2008, Server 2012
  • At least 10 GB of HDD space and 8 GB of RAM

Additional Info

Downloadable PDFs

Slide Show

  • ChemTunes∙ToxGPS®

ChemTunes and ToxGPS for Risk Assessment


Please contact us preferentially by e-mail if you are interested in evaluating ChemTunes∙ToxGPS®.