Pusan National University Researchers Revolutionize Environmental Health with Advanced Explainable Machine Learning Approach

Machine Learning with FLIT-SHAP Reveals Crucial Pollutant Interactions, Enhancing Toxicity Predictions and Environmental Health Decision-Making

BUSAN, South Korea, July 23, 2024 /PRNewswire/ -- Traditional environmental health research often focuses on the toxicity of single chemical exposures. However, in real-world situations, people are exposed to multiple pollutants simultaneously, which can interact in complex ways, potentially amplifying or diminishing their toxic effects. Conventional models that assume additive effects, like concentration addition and independent action, can be misleading in these scenarios. Although advanced statistical and machine learning methods have been employed to address this issue, they frequently fall short due to the complexity, high number of interacting pollutants and the inability to extract each pollutant's absolute effect.

To address this issue, a group of researchers led by Professor Kuk Cho from Pusan National University introduced Feature Localized Intercept Transformed-Shapley Additive Explanations (FLIT-SHAP) as a solution to these challenges. This tool is unique because it breaks down the effects of specific pollutants within a mixture, unlike traditional methods that use a broader approach. This detail-oriented technique is particularly useful for understanding the OP of particulate matter (PM) in the air, which is known to cause various health problems. This study was published in Volume 475 of the Journal of Hazardous Materials, on 14 August 2024.

"The OP of PM refers to the ability of these particles to generate reactive oxygen species, which can cause oxidative stress and inflammation in the body, leading to various health problems such as respiratory and cardiovascular diseases," explains Prof. Cho. The scientists use FLIT-SHAP to examine the OP of these particles in a mixture of pollutants. Laboratory-controlled data was used to train robust machine learning models like eXtreme Gradient Boosting (XGBoost) and their FLIT-SHAP explanation tool was used to reveal significant insights.

"We found that in a controlled laboratory environment, 55-63% of interactions were synergistic (meaning they work together to increase toxicity), while 25-42% were antagonistic (where they counteract each other). However, in real-world scenarios, antagonistic effects were more common, leading to lower overall toxicity than predicted by older models," added Prof. Cho.

One surprising finding was the significant impact of a group of chemicals called quinones, which showed more contribution to toxicity than previously thought. This insight challenges current approaches to regulating pollutants and suggests a greater need to focus on these chemicals. FLIT-SHAP not only predicted toxicity with remarkable accuracy (R² = 0.99) but also offered detailed insights into how different pollutants interact. This makes it a powerful tool for conducting realistic risk assessments and enhancing our understanding of environmental health risks.

The researchers' rigorous tests uncovered how pollutants interact at varying concentrations, revealing that traditional additive models often over- or under-estimate mixture toxicity. Through FLIT-SHAP, they pinpointed crucial nuances in pollutant contributions across different environments, underscoring the need for refined risk assessments.

While the study focused on a select group of pollutants and toxicity endpoint, further research into broader scenarios is essential. FLIT-SHAP represents a significant leap forward, offering a precise tool to evaluate chemical mixture toxicity. This advancement promises better regulatory decisions to safeguard human health effectively

Reference

Title of original paper: Machine learning-derived dose-response relationships considering interactions in mixtures: Applications to the oxidative potential of particulate matter

Journal: Journal of Hazardous Materials

DOI: https://doi.org/10.1016/j.jhazmat.2024.134864

About the institute
Website: https://www.pusan.ac.kr/eng/Main.do

Contact:
Jae-Eun Lee
82 51 510 7928
380682@email4pr.com

View original content to download multimedia:https://www.prnewswire.com/news-releases/pusan-national-university-researchers-revolutionize-environmental-health-with-advanced-explainable-machine-learning-approach-302202673.html