Project 3

Development of validated and predictive Quantitative Structure-Toxicity Relationship models that employ both chemical and biological descriptors of molecular structures and take into account genetic diversity between individuals

Description

The unique contribution of Project 3 in the context of the Center is its focus developing the universally applicable and robust predictive Quantitative Structure Activity/Toxicity Relationships (QSAR/QSTR) modeling framework which is tightly linked to toxicogenomic models and networks (Project 1) and toxico-genetic descriptors (Project 2). The modeling framework has been refined over many years of our research in the areas of QSAR methodology development and applications including novel data analytical approaches, molecular descriptors, model validation schemes, overall QSAR workflow design and proposed inclusion of multidimensional experimental end-points including toxicological, toxicogenomic and toxicogenetic data. In this project we endeavor to extend our studies and deliver tools and models that go far beyond the scope of traditional QSAR modeling by directly engaging the entire “source-to-outcome” continuum of modern experimental toxicological research. Our research entails rigorous analysis of interrelationships between chemical structure, high-throughput toxicity screening (HTS), multiple “-omics” data, genetic diversity of the organisms, and data from chronic toxicity studies. Our methodologies employ not only conventional chemical descriptors of molecular structures but also combine those descriptors with experimentally-derived biological endpoints. The ultimate goal of this project is to deliver robust modeling tools and accurate computational predictors of specific toxicity endpoints to prioritize both chemical agents and animal strains for in vitro and in vivo testing.

This project pursues the following major specific objectives:

• Develop rigorous end point toxicity predictors based on the QSAR modeling workflow and conventional chemical descriptors
• Develop novel computational toxicogenomic models based on combined chemical and biological descriptors through QSAR modeling workflow
• Develop novel computational toxicogenetic models based on combined genetic, chemical and toxicity descriptors through QSAR-like modeling workflow

These objectives synchronize with three types of computational models that in our practice contribute to the enterprise of toxico-cheminformatics. The first is more traditional and deals with the development of QSTR models using chemical descriptors for single toxicological end point data.  In this case, we develop models that correlate chemical structure (represented by its chemical descriptors) and the experimental (phenotypical, e.g., carcinogenicity) end-point data as the target property. The second deals with building models on the basis of genome-wide transcriptional profiles or multiple high-throughput biological assays (e.g., those collected in ToxCast program) that can be regarded as biological descriptors of chemical structures. These descriptors can be used in QSTR modeling either by themselves or in combination with conventional chemical descriptors. Finally, toxic effects are influenced by genetic diversity of test animals that can be formalized via genotypes. There is a challenge of developing toxicity predictors using genotypes as descriptors of mouse strains, ideally also taking chemical compound diversity into account. Such data are beginning to emerge, most importantly from the experimental work of Project 2 on this proposal, and we plan to use this data for building our models.