Dr. John Wambaugh, Environmental Protection Agency

High-throughput screening and exposure prediction tools are necessary to prioritize thousands of chemicals for public health risk. High-throughput models based on machine learning can estimate human exposure rates. Models currently exist for four sources of exposure: manufacturing, pesticide use, consumer products, and diet, with diet being the least well covered by mechanistic exposure models. Machine learning models can only predict sources of exposure when given appropriate training data. These data may be

obtained from databases or chemical surveillance of relevant media. Many of these data are limited to occurrence of chemicals in products, which is a prerequisite to, but does not guarantee exposure. To address the thousands of chemical-formulation combinations, models of migration or emission of chemicals are needed. Ultimately, exposure prediction and surveillance should allow prioritization of potential risk posed to public health