You are here:
Computational Intelligence: Building ‘Smart Models’ for Toxicology in the Era of Big-Data
Citation:
Knudsen, T. Computational Intelligence: Building ‘Smart Models’ for Toxicology in the Era of Big-Data . Presented at Society of Toxicology (SOT) 2020 Annual Meeting, Anaheim, CA, March 15 - 19, 2020.
Impact/Purpose:
This abstract was selected by the SOT to include in the Preliminary Program to highlight key events for the 2020 meeting in Anaheim, CA. It was selected for the 2020 In Vitro Toxicology Lecture and Luncheon for Students.
Description:
Technologies used by toxicologists, and the biological questions we ask, have become increasingly sophisticated with the integration of data science and computational intelligence. Computer models allow us to mine big-data and synthesize important concepts that can be applied to biologically-complex systems. To operationalize complex data for toxicological evaluation, information must be collected, organized, and assimilated into models that bridge different levels of biological organization. By reconstructing biological tissues in silico into simulated ‘virtual tissues’, we can demonstrate how dynamic changes occur in response to a particular stimuli such as biomolecular lesions introduced from real world data. This has been captured for several examples in embryo development with data from ToxCast or literature, rendering where, when, and how a particular lesion might lead to an adverse outcome facilitating mechanistic hypothesis generation and predictive toxicology. Discussion will focus on how many models are needed, how smart models must be to support decision-making in the animal-free (3Rs) zone, and practical considerations for technology development, application, and training for predictive toxicology. Disclaimer: does not reflect EPA policy.