Service Communautaire d'Information sur la Recherche et le Développement - CORDIS


DEMETRA Résumé de rapport

Project ID: QLK5-CT-2002-00691
Financé au titre de: FP5-LIFE QUALITY
Pays: Italy

A combined QSAR model for pesticides to predict acute toxicity for rainbow trout (Oncorhynchus myskiss)

The result "combined QSAR model for pesticides to predict Acute Toxicity for Rainbow Trout (Oncorhynchus myskiss) 38655" is given in the form of a fully predictive QSAR model built by means the combination of different individual QSAR models. The trout hybrid model has been obtained with GMDH (a self-organizative methodology of function fitting). Inputs for this model were a Partial Least Squares and two Artificial Neural Networks individual models. The basic idea of the hybrid model, which merges different points of view and combines the capabilities of different algorithms, is that different models can be more or less powerful in one or another aspect, and that they can be improved by combining positive performances of individual models. The target of the combined model is to cover as much as possible the space of the chemicals it has to model, reducing the mistakes.

These individual models were built using 229 pesticides as training set whose toxicity values were obtained from the more trustworthy pesticide databases (see TIP corresponding to Data of pesticide toxicity, structures and chemical descriptors: Acute Toxicity for Rainbow Trout (Oncorhynchus myskiss) 38662).

This result is implemented as a part of a Java applet, also able to run as standalone version, whose only input is the molecular descriptors of the new pesticide to be predicted.

The present result fulfils completely the quality checkpoints marked by OECD and offers a good alternative for supporting the progressive reduction of animal experimentation covering the lack of proper QSAR models for regulatory purposes in the pesticide risk assessment. Being one of the first successful approaches in this area, this model currently has an advantageous position to successfully start to be used by regulatory bodies.

The combined QSAR model for Rainbow Trout is in line with the recent EU legislative initiative REACH on the possible use of QSAR models as an alternative animal experimentation.

The main idea that underlies in the whole development of this model is creating a tool to be used (not only formally, but in practice) within the Pesticide Directive 91/414, thus a crucial point of the project involved a deep evaluation of criteria for its use.
QSAR models, eventually evaluated for regulatory purposes, that have been developed so far use common QSAR criteria not specific for regulatory purposes.
This places the present model in an advantageous position due to its correct timing to the legislation, and makes of it a desirable tool for all the agents involved in risk assessment.

The major user category we have identified within the DEMETRA project is regulatory body.

Non-governmental agencies are also major potential users of the models, for issues similar to those of regulators.

Another important category is industry in the field of plan protection products.
Another major category of users is academy and research institutes.
The first steps in result dissemination has been already taken organizing an European workshop. Members of key institutions have been invited to this meeting covering most of the regulatory bodies and industries dealing with pesticides at European level.
From the ethical point of view this result can help to reduce the suffering derived to the exposure of rainbow trouts to non tested compounds. The use of our free model is cheaper than performing animal experimentation and there is a social concern that can be fulfilled by our model.

It helps to reduce the time in assessing the environmental risk of new pesticides/chemicals, the budget foreseen for experimentation and the material, energy and space necessary to perform the experimental assays. Additionally these animal assays need big infrastructures since it is necessary to store big tanks refilled with fresh water and with high amounts of energy for maintenance of oxygen and temperature. Since the QSAR model reduces the number of experimental tests this results in a significance reduction of the expenses.

Additionally this model can help to boost the research in new pesticides. This model allows the evaluation of environmental risks in a prognostic way, therefore it is not necessary to perform the syntheses of the compound. This fact allows to the industries to deal with a higher amount of chemicals candidates to become a pesticide in the market. Also it might help to evaluate metabolites, allowing a more precise knowledge of the mechanistic features of the environmental risks.


Emilio BENFENATI, (Laboratory Head)
Tél.: +39-0239014420
Fax: +39-0239014735