A new technique can more accurately detect the presence of nondegradable, synthetic chemicals that linger invisibly in ecosystems thanks to their as-yet unknown structures. These aptly nicknamed "forever chemicals" persist and accumulate in the environment, causing cancers and developmental disorders across all organisms.

When tested on wastewater samples collected in 2011 from a Chinese industrial park, the framework identified 31 classes of these forever chemicals, which are also known as per- and polyfluoroalkyl substances (PFAS). Within those 31 classes were 17 classes of PFAS that had gone unreported until now.

"[This] reveals a greater presence of these chemicals in the environment than previously known," said Si Wei, a specialist in environmental chemistry, professor at Nanjing University and corresponding author of the research published in Science Advances. "[Our] finding is critical for understanding the potential impact of PFAS on the environment and human health."

During the mid-20 th century, PFAS appeared on the scene and rapidly gained popularity because of their innate resistance to heat, water and oil. These qualities made them perfect for nonstick pans, waterproof fabrics, fire-fighting foam, food packaging and more. However, as the 21 st century began, national agencies including the Centers for Disease Control and Prevention and international authorities noticed that the substances' ubiquity had led to unexpected side effects: PFAS were furtively entering water sources, land and animals, and they were causing cancer, reproductive defects and more.

Like many drugs that break down in the gut and then disperse throughout the body, PFAS too break down into compounds, or "seeds," as they slink through ecosystems.

"These seeds act as reference points. Once we identify a seed, we can use it to find other PFAS that have similar structures," said Wei. "It's like finding a particular tree in a forest. Once you recognize that tree, you can easily spot others of the same type."

This tactic can provide health and environmental policymakers with information necessary for taking regulatory action against these pollutants. Just this spring, the U.S. Environmental Protection Agency passed legislation to address PFAS contamination. Yet, a problem remains. Companies are inventing different PFAS that may circumvent legislative efforts, and most do not have to disclose any information about these newer chemicals' design.

"New PFAS are kind of like cousins to the old ones they share similar properties and similar structural features, but [are] not completely consistent," said Wei. "Because they're different, they're not as familiar to scientists and regulators, which makes it more difficult to keep track of them and evaluate their potential risks."

Many new PFAS are so different in structure from their predecessors that they and their seeds are invisible to existing tools.

Noting this clear need for new investigatory technology, Wei and his colleagues developed a platform, called APP-ID, that combines machine learning algorithms with a high-resolution mass spectroscopy molecular network approach. Researchers conventionally use this latter strategy to screen for microbial natural products and metabolites with therapeutic potential that lurk in the deep ocean's hydrothermal vents.

The approaches "cluster structure similar compounds and identify the unknowns based on the information of knowns," said Wei. "By applying molecular networking, we can map out the relationships between known and unknown PFAS."

In tests, the PFAS detection framework discovered unknown chemicals with 58.3% accuracy an improvement over three other current methods with 43.8%, 37.5% and 12.5% accuracy, respectively.

The team also had APP-ID evaluate wastewater samples taken over a decade ago from a fluorochemical industrial site in China. It successfully unearthed 733 PFAS belonging to 31 classes. Roughly 54% of those classes seen had never been described before. Notably, 10 of the unreported classes were made of single compounds, which are particularly hard to trace using traditional methods.

Next, the group had the tool retrospectively screen a public repository called MassIVE. This databank is renowned among scientists because it holds 15,000 datasets with environmental and human data from 50 countries. When analyzing a variety of environmental and human samples from 20 of those countries, APP-ID exposed 126 PFAS comprised of 81 unknown, eight legacy, and 37 emergent or newer, already-characterized types of PFAS chemicals. Essentially, 64% of the 126 had never been catalogued before a result that underscores how newer generations of forever chemicals are hiding in plain sight.

Wei hopes that APP-ID will aid future efforts to conceptualize unknown PFAS accumulation globally. "It's reasonable to assume that the variety of PFAS has expanded over the past decade," he added. "Further research is necessary to uncover the historical trends and forecast future patterns of both known and unknown PFAS."

Read the rest here:
Machine Learning Approach Uncovers Unreported PFAS in Industrial Wastewater | American Association for the ... - AAAS