Per and polyfluoroalkyl substances (PFAS) are a widely manufactured class of chemicals with unique physicochemical properties and concerning degradation profiles in terms of environmental persistence and long-range transport. Bioconcentration factor (BCF) is the ratio of concentration of chemical in fish to its concentration in water; BCF is an important metric of bioaccumulation for regulators. The purpose of this study was to assemble experimental data from online data sources and literature to develop QSAR (quantitative structure activity relationship) models to predict bioconcentration factor for PFAS compounds. Using Java programming, the data was filtered on relevant BCF metadata. For example, only records where the whole body tissue was measured were kept. Records were stored in a PostgreSQL database using a consistent data format. Each record was mapped to a unique substance ID in EPA’s Distributed Structure-Searchable Toxicology Database (DSSTOX). The substance ID allows one to associate each record with a “QSAR-ready” SMILES string which is then used to generate molecular descriptors. Data set records consist of an ID value, a log<sub>10</sub> BCF value, and the molecular descriptor values. Discordant records were omitted and the data sets were randomly split into training and prediction sets. Models were built using machine learning libraries available in Python including random forest and support vector machines (SVM). Some models were constructed using data from all chemical classes (global model) and others using only PFAS compounds (local model). Results for PFAS in the test set were compared for the global and local models.

Development of PFAS Bioconcentration models using online data sources and python-based machine learning

Fall 2022 abstracts, posters, and presentations will launch in mid-September.

ACS Fall 2022

The official archive and repository for all ACS Meetings & Expositions abstracts and select posters or presentations.