BGSU researchers use machine learning to study 2023 East Palestine, Ohio, train derailment

Environmental toxicologist Dr. James Metcalf leading project to better understand human costs from chemical spill in East Palestine, Ohio

Two years after a train derailment exposed an Ohio community to toxic chemicals, the Ohio Attorney General’s office is leveraging the expertise of researchers at Bowling Green State University, who are using machine learning to help discern possible long-term health consequences of the incident.

Dr. James Metcalf, an environmental toxicologist and the Hollingsworth Smith Endowed Chair in Biological Sciences at BGSU, is spearheading the research project that will use artificial intelligence to examine the human costs stemming from the 2023 Norfolk Southern train derailment in East Palestine, Ohio

The project has multiple innovative wrinkles that seek to definitively understand whether the chemicals released – primarily vinyl chloride, a precursor to PVC piping that is poisonous to humans – negatively affected the long-term health of people exposed.

The effects could take decades to fully understand, but scientists are acting now to build a data set that could prove invaluable in future years.

“The idea is to establish baseline levels of disease in the population so in the future we can go back and interrogate this data to see if there have been changes,” Metcalf said. “The major reason to do this now is that if, in the future, something were to appear in the population, the methods are already in place to do the analysis quickly and accurately to give the people of East Palestine the medical treatment and support they may need.”

Funded by a state grant of $50,000, the idea for the project came through the attorney general’s environmental advisory council, which is made up of a dozen experts, including Metcalf.

Dr. Jon Sprague, a BCI Eminent Scholar at BGSU and the Ohio Attorney General’s Center for the Future of Forensic Science, said the council was perfectly suited to undertake a research project on the topic.

“It really came together naturally,” Sprague said. “The environmental advisory council was already meeting regularly and following current events in Ohio that related to environmental situations that the attorney general’s office should be aware of, and obviously the train derailment had a major impact.

“Each of those individuals on the council brings a unique skill set to our team, and James’ skill set matched exactly what we needed for this type of project.”

Making a determination on whether the exposure is influencing human health has its challenges: adverse effects might not appear for decades, the exposure’s impact could be subtle and the overarching data is subject to human bias.

To address these challenges, scientists are taking a longitudinal approach by first establishing baselines and comparative groups – providing critical data points against which future findings can be compared – and using machine learning in data analysis to remove the possibility of bias.

“If there is an adverse outcome, this would likely be very subtle, perhaps a slight increase in certain types of cancers,” Metcalf said. “That may be very difficult to discern accurately by the human eye because humans evolved by looking for patterns in things. By using machine learning, we remove that inherent bias to look for patterns that may or may not be there.”

BGSU graduate student Luke Robbins ’23, who had a chance to try research as an undergraduate in Metcalf's lab, will be assisting with the high-profile project while he is obtaining his graduate degree in biological sciences at BGSU, which the Wall Street Journal ranked as the No. 1 public university in Ohio for student experience.

“As I got to know lab work more, it helped me flesh out what I really wanted to do," Robbins said.

At its core, the project seeks to understand a topic of great importance to citizens of Ohio, but the research could have an impact that goes far beyond the University’s home state.

Understanding the health consequences of the train derailment could influence policy all over the world, Sprague said.

“If we do this right and use machine learning to determine the health consequences of the train derailment, this could be applied not only to Ohio, but across the country and the world,” Sprague said. “This has major impact potential, and not just in the state of Ohio.”