iSafeRat® (in silico Algorithms For Environmental Risk And Toxicity) is a registered trade mark and the name given to the toolbox of High Accuracy QSAR (HA-QSAR) modules produced by KREATiS. Initially designed to meet the REACH requirements to fulfil endpoints of regulatory dossiers, iSafeRat® aims to be the most reliable and accurate in silico approach to replace all kinds of experimental studies related to human health and the environment.
The structure (SMILES) is used to predict critical PC properties (log Kow and solubility) which are themselves used to predict ecotoxicity and human health endpoints.
Potential for endocrine disruptor modalities is currently predicted in-house.
Our objective is to provide an endpoint value with an accuracy at least as good as that obtained by the best experimental technique available for that endpoint following OECD Guideline methods but for a fraction of the price of a laboratory study.
Each endpoint value that you order will be provided in a full report format containing a project and study number, information on the substance provided by the client, a brief explanation of how the prediction was made, the result and confidence limits and in Annex a full QSAR Model Report (QMRF) and QSAR Prediction Report (QPRF) necessary for successful regulatory submission.
The toolbox uses coded chemical structures (SMILES) of organic chemicals to:
► predict critical physico-chemical parameters
► define structural alerts within the chemical structure and fit them within a highly structured classification scheme (Mechanisms of Toxic Action or MechoAs)
► from the physico-chemical parameters and the MechoA, quantify their (eco)toxicity
► Applicability domain information* to immediately provide upfront information on the regulatory acceptance status of the prediction
► Confidence interval calculations allowing you to judge on the models performance
*3 options are proposed, “inside applicability domain”, “outside applicability domain” or “extrapolated”. “Extrapolated” means we believe in the result but statistically it is not currently within domain. The prediction can still be very useful.