Novel algorithms in prediction of outcomes of exposure to environmental chemical mixtures
Summary
Rapid technological developments have allowed for the identification or development of over 50 million different chemicals, the vast majority of which emerged in the last two decades. The speed of new chemical production or isolation intensifies with geometric progression; while 33 years were needed to isolate the first 10 million chemicals, it took fewer than nine months for the last 10 million. It is estimated that a new substance is isolated or synthesized every 2.6 seconds. The current problems in chemical safety assessment (before their rapid dissemination on the market) are underscored by a very low capacity for new chemical testing (only ~300 chemicals and pesticides can be tested annually in the U.S.), development of new classes of chemical compounds with mostly unknown toxicological profiles (i.e., nanoparticles), and the necessity of development of novel endpoints for chemicals with non-genotoxic modes of action, such as endocrine disruptors (i.e., pesticides). However, the largest problem is understanding and predicting the toxicity of mixtures of chemicals that are ubiquitously present in the environment.
Current paradigms in chemical risk assessment are built on evaluation of the individual component and are based on regulatory test guidelines developed to provide a dose-response assessment that estimates a point of departure. The latter has traditionally been used to extrapolate the quantity of a given substance above which adverse effects can be expected in humans. This approach, however, does not consider exposure to the unusual and heretofore untested combinations of chemicals. The combined pharmacological and toxicological effects of mixtures are often unknown or can differ markedly when compared to their exposure as single entities. This may result in unanticipated adverse effects, usually not caused by exposure to individual compounds alone.
Therefore, we propose to develop a framework for identification of the effects of chemical mixtures. This approach will be based on utilization of artificial intelligence that will utilize the open databases (i.e., PubMed, PubChem) for machine learning and developing novel algorithms that will be able to effectively predict the toxicity potential for a mixture of given chemicals. We are confident this product will be of particular importance to regulatory agencies (i.e., U.S. EPA and FDA), as well as private industry where it will be effectively utilized during the process of novel product development.
Keywords:
- artificial intelligence
Researchers:
- Igor Koturbash (Author)
- Justin Zhan
- Gunnar Boysen