Gateway for:

Member Countries

Models of Toxic Action of Chemicals


Development and Application of QSAR Rodent Toxicity Models for Chemical Compounds

Tech Area / Field

  • ENV-MRA/Modelling and Risk Assessment/Environment

8 Project completed

Registration date

Completion date

Senior Project Manager
Rudneva V Ya

Leading Institute
Institute of Physiologically Active Substances, Russia, Moscow reg., Chernogolovka

Supporting institutes

  • Institute of Biomedical Chemistry, Russia, Moscow


  • University of Liverpool, UK, Liverpool\nEuropean Commission / Joint Research Center / Institute for Health and Consumer Protection, Italy, Ispra

Project summary

Chemical Abstracts Service contains information over 28 million chemicals. About 200,000 different chemicals are produced commercially and utilized every year and 2000 to 3000 new chemicals are added annually to that list. Current experimental methods do not provide a possibility to study toxic and side effects for all synthesized compounds. Complete toxicological data, are available only for less than 10,000 of them. At the same time, study of side and toxic effects becomes more and more important in development of new medicines and application of known drugs. According to the nowadays data, toxic and side effects of drug-candidates are the reasons of 50% failures in drug development. The experimental study of toxic and side effects is time-consuming and highly expensive. The reason of that is the complexity and multifactor nature of such phenomenon as “toxicity” since the initial molecular mechanisms of toxic effects are perse and the most part of experiments are performed in experimental animals. Along with ethical aspects such experiments cause criticism owing to insufficient validity of extrapolation of obtained data on human. Another limitation of such approach is impossibility of its use at the stage of project planning when the methods of synthesis have not been developed and sufficient amount of biological active substances is not available yet as samples for experimental testing. The new European Society initiative (REACH - The new European Chemicals regulation), that will start at June of 2007, anticipates development of computer-aided methods for analysis of “structure-activity” relationships (e.g. IUCLID) and study of toxic effects for 30.000 chemical substances. The registration process requires the manufacturers and importers to generate data for all chemicals substances produced or imported into the EU above one tone per year. REACH is a unique basis for development and validation of algorithms for environmental risk assessment of chemical substances. It will increase considerably a number of experimental data on toxicity of chemical substances and will stimulate both utilization of available methods and development of new computer-aided methods based on Quantitative Structure-Activity Relationships (QSAR) analysis.

QSAR models can help in estimating of effects of chemicals with lack of experimental data on human health. They use the physical and chemical properties of a given chemical and other properties such as the electronic attributes embedded in a chemical structure. These properties are either easily available in the literature or can be calculated using quantum chemical methods and other established QSARs for each property. Since QSAR/QSPR models are dependent on physicochemical properties alone, QSARs are in a position to estimate potential health effects of current as well as new or modified chemical agents. In addition, the time and resources required to generate a hazard estimate using QSARs is a fraction of the time and resources required to assess the health effects using conventional empirical testing.

Therefore, the objective of this project is to develop new QSAR/QSPR (Quantitative Structure-Properties Relationships) models in the field of rodents’ toxicity, validate these models and estimate their applicability for risk assessment. To achieve this purpose, the following tasks will be solved:

  • Identification of biological end points and modes of toxic action of chemical compounds that are important for chemical risk assessment.
  • Creation of the training and validation sets for further QSAR/QSPR studies.
  • Determination of physical-chemical parameters and structural descriptors important for creation of QSPR models of toxicity.
  • Creation of robust predictable models of toxicity of chemicals to rodents on the basis of QSPR analysis for different classes of compounds and various biological end points.
  • Creation of robust predictable models of toxicity of chemicals to rodents on the basis of QSAR/SAR analysis for different classes of compounds with different mechanisms of toxic action.
  • Validation of the created models and evaluation of their applicability to chemical risk assessment.

An innovative QSAR/QSPR approaches developed during the past twenty years by two Russian research teams leaded by Prof. Oleg Raevsky (IPAC) and Prof. Vladimir Poroikov (IBMC) will be combined with the information available from public and commercially available databases and literature. It is suggested that QSAR models, developed on the basis of common mode of action (MOA) estimation, will be more robust and predictive than those based on belonging to a particular chemical class. More stable and predictive models of rodent toxicity can be created using original combination of Similarity and QSAR approaches, recently proposed by Raevsky. Complementary to the classic QSAR analysis based on physical-chemical descriptors, PASS (Prediction of Activity Spectra for Substances) method can be applied for creation of models. Since PASS estimates chemical compounds in biological space it would be possible to apply statistical clustering techniques to the output predictions for a large number of potential targets and see what types of groupings will be obtained. This gives the answer whether the predicted biological activities can be used to place chemicals into groupings based on anticipated modes of action, and how are these groupings comparable to other methods of clustering that do not use much biologically based information. Moreover, MNA descriptors and mathematical algorithm used in PASS are shown to provide a general approach for analysis of different QSAR/QSPR problems. Since PASS allows addition of new compounds and new kinds of biological activity to its training set, the program can be re-trained using data available from public and commercially available databases and literature, and quality of the models will be evaluated during the training procedure by leave one out cross-validation.

Validation of the models that will be developed in the framework of the project will be performed both in leave one out cross-validation and in predictions for independent test sets. Applicability area for each model will be determined, as well as recommendations for their use will be prepared.

Keeping in mind an innovative features of approaches that will be utilized and further developed in the framework of the project, one may expect that the implementation of the project results will improve significantly the quality of risk assessments and, as a consequence, decreasing of chemical risks for human health and increasing the quality of the environment.