QUANTUM-TOX - Revolutionizing Computational Toxicology with Electronic Structure Descriptors and Artificial Intelligence

Abstract

Toxicology is at a crossroads. With ever more drugs going to market and more chemicals having an environmental impact, the need for fast, cheap and accurate technologies to assess toxic effects is pressing. Computational toxicology provides an array of tools and methods for toxicity prediction only using computer approaches. Conceptually, computational toxicology has significant advantages since testing is fast and cheaper than in vitro. However, currently computational toxicology has severe limitations. Predictions typically use Quantitative Structure-Activity Relationship (QSAR) models that rely on large sets of molecular descriptors. This causes severe problems since the methodologies cannot assess chemicals different than the ones used to develop the QSAR models, and when that is possible, the very large number of descriptors limits understandability. Therefore, new methodologies are needed to address those shortcomings. This project will develop a new type of descriptor, totally based on quantum mechanics that can cover the whole chemical space and relies on a small number of parameters that are easily interpretable. Starting with meaningful chemical perturbations, that extract the behaviour of the chemicals in assumed mechanisms of toxic action, the approach will develop specific Electronic SIGNatures (ESigns). ESigns are mathematical invariants that map the results of the quantum chemical calculations. Using Artificial Intelligence, the ESigns will relate to toxicity. The new approach introduces a momentous change in computational toxicology. It can cover the whole chemical space, since we are abandoning predictions based on molecular structures, uses fewer parameters, and can be related to the new trends in toxicology regarding use of pathways information. In fact, it is a powerful tool to allow accurate toxicology predictions solely on the basis of biochemical and chemical insight.

Lead Participant

Project Cost

Grant Offer

THE UNIVERSITY OF MANCHESTER £528,846 £ 528,846
 

Participant

INNOVATE UK

Publications

10 25 50