You are here:
Evaluation of High-throughput Chemical Functional Use Models (ISES Annual Meeting)
Citation:
Phillips, K., K. Isaacs, D. Chang, AND K. Markey. Evaluation of High-throughput Chemical Functional Use Models (ISES Annual Meeting). ISES-ISEE 2018 Joint Annual Meeting, Ottawa, Ontario, CANADA, August 26 - 30, 2018.
Impact/Purpose:
This presentation will cover how well models that predict why a chemical is in a consumer product does with a set of data that provide the manufacturer's reason for a chemical being in a product or industrial process. This data is provided to the EPA and is made publically available under the Chemical Data Reporting (CDR) Rule.
Description:
EPA’s Office of Research and Development has generated high-throughput quantitative structure-use relationship (QSUR) models which use the structure of chemicals to predict the end role (functional use) a chemical could play in consumer products and chemical processes. As functional use provides clues on the intentionality of a chemical’s addition to a product, as well as estimated concentrations, prediction of functional uses for data-poor chemicals can provide support to those seeking to rapidly prioritize thousands of chemicals for risk and alternatives assessments. The QSUR models were built upon a training set of functional use data derived primarily from consumer product information sources. Here, we evaluated the functional use predictions of the QSUR models with publicly available functional uses reported by manufacturers under the Chemical Data Reporting (CDR) Rule of the Toxic Substances Control Act. A total of 5153 chemicals in CDR had reported function information that could be compared to model predictions. However, only 2400 compounds in CDR had an available structure. Of those 2400, 1581 had predictions with a probability greater than 80% that were within a model’s domain of applicability. In addition, we used reported CDR use sector (industrial or commercial) to explore structural differences in chemicals having the same function in different use sectors to assess the applicability of models trained on primarily consumer products to compounds in industrial applications. These comparisons will drive the development of improved models. The views expressed in this abstract are those of the authors and do not necessarily reflect the views or policies of the US Environmental Agency.
