research

In medicinal chemistry, environmental toxicology, and human health hazard assessment of chemicals, there are many thousands of chemicals that need to be tested for their effects on human and ecological health. Most of them don't have much experimental data for the estimation of their potential hazard. Therefore, there is a compelling need for data on various properties of chemicals. Computational and theoretical methods, such as quantitative structure-activity relationships (QSARs), rely on this problem in a rapid and cost effective way. Indeed, the main paradigm of medicinal chemistry and modern toxicology research is that biological activity, as well as toxicological, physicochemical, and environmental properties of compounds is inherent on molecular structure. Quantitative-structure activity relationships are mathematical models that relate experimental properties of molecules to their structure. Since chemical structures are non-quantitative entities, they are before translated into numerical values by using molecular descriptors. The molecular descriptor is the final result of a logic and mathematical procedure which transforms chemical information encoded within a symbolic representation of a molecule into a useful number.

As a result of the incoming operative step of a legislative proposal, adopted by the European Commission, for a new chemical management system called REACH (Registration, Evaluation and Authorisation of Chemicals), it is expected that the use of QSARs for regulatory purposes will increase. In effect, they are valid alternative methods for predicting the environmental and toxicological properties of chemicals, in the interests of animal welfare but also of time-effectiveness and cost-effectiveness. Research in QSAR has advanced rapidly in the last decades and accordingly a large variety of models have been proposed, each based on different molecular representation and/or theoretical and computational methods. Nevertheless, we still suffer from several problems leading to a scarce concrete use of QSAR models. Most of these problems are due to the difficulties in sharing and applying models even within the scientific community, often for the lack of fundamental information needed to understand and correctly apply them. Most of the published QSAR models are not reproducible because of ambiguous algorithm, not well-defined molecular descriptors, lack of training set data, etc. Moreover, some software exist for molecular property estimation, but predictions are often provided for not many endpoints and the implemented models are frequently black-boxes, since they do not provide users with the suitable information for evaluating estimate reliability. Only in the last few years, some efforts have been done to define the necessary information, which should be associated to a QSAR model for its validation for regulatory purposes, and to implement valid QSAR models in specialized toolboxes, such as the QSAR Application ToolBox (under development by OECD).

The present project aims to support and strengthen the done work until time in the implementation of QSAR toolboxes by proposing a novel open computer format for QSAR model information storage based on XML (eXtensible Markup Language). This format will be promoted within the scientific community with the ambition of affirming it as the standard format for QSAR model information exchange. The need of a standard computer representation of QSAR models is felt today more than ever because only by computer-based tools information can be quickly spread through Internet and managed by software. There are thousands of models proposed and published in the literature. They are based on statistical and mathematical approaches, which are well-suited to automation via computer-based tools. The computer format for QSAR models resembles to some extent the chemical structure files for molecules, which allow molecules to be managed in databases and processed by software for descriptor calculation, property estimation, similarity analysis, etc. By the proposed computer format, existing and future models may be easily collected in specialized databases and retrieved by keywords for research purposes and automated QSAR toolbox implementation. Moreover, by means of an electronic collection of models, easy comparison will be performed in order to find out discrepancies among property estimates and highlight which endpoints and/or classes of compounds require further efforts in QSAR research.

The impact of this project is expected to be relevant, as it provides tools to boost QSAR modelling research and promotion, along with a concrete answer to the need of suitable, affordable and ethical tools for human and environmental risks assessment. The project outcome will support the implementation of EU policies related to risk assessment of human health and environmental safety, since industry and regulatory authorities will be able to find and use tools for evaluating the endpoints of interest and eventually adapt them to specialized toolboxes. From an ethical point of view, the proposed chemoinformatics tools will lead to the promotion of QSAR models as useful methods not involving animal testing to handle with data gap and even more important they will allow a concrete use of these methods. More QSAR tools will be accepted by regulatory authorities and adopted by industry, more animal lives will be saved by switching from in-vivo or in-vitro methods to in-silico prediction of chemical properties.

Milano Chemometrics and QSAR Research Group
Department of Environmental Sciences
University of Milano-Bicocca