Big Data-Enabled Modeling Approaches to Address Challenges in the Chemical Exposure to Human Risk Continuum

Abstract

The concern of human health risks from exposure to chemicals has grown which has resulted in numerous studies in experimental animals to assess the risk. There is a significant gap in data and models as they often fail to address uncertainties, population variability, and complex interactions between chemicals in a mixture. Also, most experiments have focused on the effect of a single compound for a certain duration; however, exposure to environmental chemicals is not commonly limited to one single chemical or one person at a single time point. These data gaps and the absence of appropriate models may threaten public health. To overcome these problems, computational toxicology is a rapidly growing tool to determine the accurate dose-response relationship for the whole population and characterize hazards using statistical modeling for risk assessment. Utilizing novel probabilistic modeling for exposure estimation and hazard characterization enables us to address these limitations through the integration of real-world big data to facilitate the estimation of possible hazards in the population. In order to test the central hypothesis that the big data-enabled models result in a significant improvement in addressing risks of population, we will pursue three specific aims: 1) Characterize chemical spatial and temporal trends using “exposure big data” after an industrial accident; 2) Use Bayesian concentration addition models to estimate population variability in effects of chemical mixtures by using “in vitro big data” from a human population in vitro model; and 3) Build a Bayesian model with “animal bioassay big data” to characterize the population and individual cancer risks from chemical exposures. The accomplishment of the aims will demonstrate how a variety of “big data” types and approaches can facilitate the estimation of risks after an industrial disaster, human variability in responses to chemical mixtures, and probabilistic population cancer risks. Eventually, we expect that the development of computational research methodology will fill the critical gaps in toxicological data and improve the feasibility and accuracy of human health risk assessment.