Servicio de Información Comunitario sobre Investigación y Desarrollo - CORDIS

The development of models (QSARs) for the prediction of the endocrine disruption capability of compounds. and (QAAR) for the prediction of the activity to one test/animal

There is a significant requirement to predict biological activity from chemical structure. There are various methods to achieve this, all loosely defined as quantitative structure-activity relationships (QSARs). Such tools have proven successful in priority setting for risk assessment and are likely to gain some form of acceptance in the regulatory community. QSAR modelling employs statistical approaches to correlate or rationalise variations in the biological activity of a series of chemicals with variations in molecular structure.

In this study a large database of quantitative and qualitative data have been collated for oestrogenicity and two different QSAR approaches were applied to enable prediction from structure. The first approach is introduction of simple molecular features as “rejection filters” to eliminate chemicals that are unlikely to be oestrogenic. The rejection filters should not generate any false negatives and be able to reduce significantly the number of chemicals for further evaluation.

The dataset used in this study consisted of over 500 chemicals for which at least one oestrogenicity endpoint was available. Following molecular features were studied: molecular weight range, lack of a ring motif, log P range, number of hydrogen bond donor and acceptor groups and number of flexible bonds. It was founds that molecular weight range together with the lack of a ring motif were successful criteria, eliminating larger number of inactive chemicals with no false negatives. The second QSAR approach involved a linear correlation between oestrogen receptor binding affinity of 131 chemicals and the structural descriptors. The resulting QSAR was able to predict log RBA values of chemicals in the test set with a good accuracy.

In order to study oestrogenic activity more thoroughly, three-dimensional (3D) quantitative structure-activity relationship (QSAR) and structure-activity relationship (SAR) analyses were applied concurrently to a data set of highly selective oestrogen receptor (ER) alpha and beta agonists.

The data set consisted of diphenolic azoles characterized by similar structural skeletons but with different binding modes to the oestrogen receptor site. Models were developed separately with respect to the relative binding affinities (RBAs) to ER alpha and ER beta. Steric and electrostatic fields were calculated for a training set of 72 compounds using comparative molecular field analysis (CoMFA). The model developed for ER alpha RBA yielded a squared correlation coefficient of 0.91 and a cross-validated squared correlation coefficient of 0.60. The model developed for ER beta RBA yielded a squared correlation coefficient of 0.95 and a cross-validated squared correlation coefficient of 0.40. Both models were validated successfully using an external test set of 32 compounds.

A new concept of test set evaluation based on the variability of the biological response due to the variability of the living organism was been introduced. The CoMFA analysis was supported by a SAR study. In addition to the most favourable steric and electrostatic regions identified by CoMFA, a number of structural descriptors were identified as being important for binding.

Quantitative activity-activity relationships were also investigated to determine the relationship between biological assays. The measurements of oestrogen disruptor chemicals resulting from the [3H]-estradiol radioligand binding assay, reporter gene assay, and E-screen assay for determination of the estrogenicity were evaluated with respect to the correlation between the endpoint values and the concordance between the assays. The correlation between the data points was determined using regression analysis. The concordance between the assays was determined by the estimation of the statistical equality of two populations using the Kolmogorov-Smirnov test.

Experiments were carried out for the following cases:
- For endpoints of compounds for which identical or different assays were performed in different laboratories and
- For endpoints of compounds for which all assays were performed in the same laboratory.

The results demonstrate that for a small number of measurements (17-21 datapoints), the endpoint pKi shows high correlation to the RBA (derived from IC50 or EC50), RA and PC50 endpoints and agonism. The Kolmogorov-Smirnov test resulted in no evidence of an existing difference between the assays applied to the endpoint pKi and the endpoints RBA, RA and PC50, respectively for the same number of data points.

The significance of the QSAR and QAAR analysis is that they provide methods to ascertain the hazard associated with a chemical without resort to costly and time-consuming animal testing.

Información relacionada


Mark CRONIN, (Professor of Predictive Toxicology)
Tel.: +44-1512312402
Fax: +44-1512312170
Correo electrónico
Síganos en: RSS Facebook Twitter YouTube Gestionado por la Oficina de Publicaciones de la UE Arriba