: The objective of this study is to develop machine learning models to estimate oxidative potential (OP) of particulate matter (PM), particularly PM2.5, from its chemical composition. PM toxicity originates from the excess generation of reactive organic species (ROS) over the antioxidant capacity in cells, which can oxidize cellular components. The oxidative stress has been found to be more strongly linked with the health effects of PM compared to routinely measured PM mass. The capability of PM to induce this stress through generation of ROS is termed as oxidative potential (OP). In this research, we will generate the training datasets by measuring OP endpoints of different solutions containing varying concentrations of primary organic aerosols (POA) such as Ammonium Sulfate (NH4SO4) and Ammonium Nitrate (NH4NO3), Humic-like Substances (HULIS), Hydrocarbon-like and Oxygenated Organic Aerosols (HoA) and metals [Cu(II), Mn(II) and Fe(II)] found in PM2.5. The test concentrations of model species will be chosen based on their typical ambient concentrations and conventional filter extraction protocol. The OP will be measured using three end points: dithiothreitol assay (OPDTT), consumption of reduced glutathione or GSH (OPGSH) and hydroxyl radical (OH) generation rate in surrogate lung fluid or SLF (OPOH-SLF). A total of 648 (3 concentrations of NH4SO4, NH4NO3, HULIS and HoA and 2 concentrations of Cu(II), Mn(II) and Fe(II)) samples will be analyzed to simulate all possible combinations. First, binary mixtures will be used to quantify any synergistic or antagonistic interaction between the components, followed by tertiary or quaternary mixtures. The measurements will be carried out using a Semi-Automated Multi-Endpoint ROS Activity Analyzer (SAMERA). The experimental data will be modeled using five machine learning models: Support Vector Machine, Artificial Neural Network, Random Forest, Gaussian Process and XGBoost. The models will be validated against OP measurements of ambient PM2.5 samples collected from different locations in the Midwest USA. The most suitable model will be selected based on Coefficient of Determination (R2) and Root Mean Square Error (RMSE). Our final goal is to integrate the model with GEOS-Chem for PM2.5 OP forecast across Midwestern US.

Application of machine learning in developing models to estimate the oxidative potential of particulate matter from its chemical composition across the midwestern US

Where thousands of chemistry professionals meet to share ideas and advance scientific and technical knowledge. Spring 2023 abstracts, posters, and presentations will launch at the end of March.

ACS Spring 2023

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