Identifying priority PBT-like compounds from emerging PFAS by nontargeted analysis and machine learning models.

Area of Science:

• Environmental Chemistry
• Toxicology
• Computational Chemistry

Background:

• Traditional per- and polyfluoroalkyl substances (PFAS) are being replaced by emerging PFAS with largely unknown environmental and health properties.
• Assessing the persistence, bioaccumulation, and toxicity (PBT) of these new chemicals is crucial for environmental protection and public health.

Purpose of the Study:

• To compile a comprehensive database of emerging PFAS.
• To evaluate the PBT properties of emerging PFAS using advanced machine learning models.
• To identify high-priority emerging PFAS for regulatory and industrial attention.

Main Methods:

• Nontargeted analysis was employed to identify emerging PFAS.
• A database of emerging PFAS was created.
• Machine learning models, including graph attention networks, Insubria PBT Index, EAS-E Suite, VEGA, and ProTox-II, were utilized for PBT property prediction.
• Chemical vendor and patent databases were searched for prioritized PFAS.

Main Results:

• A total of 282 emerging PFAS homologues (21.8%) were predicted as PBT.
• 140 PBT PFAS were detected in industrial and environmental samples, categorized into four main groups.
• 10.1% of prioritized PFAS were linked to chemical vendors and 19.6% to patents.
• Aqueous film-forming foams and fluorochemical factories were identified as primary sources.

Conclusions:

• The developed database and PBT screening methodology provide critical data for regulatory bodies like the US Toxic Substances Control Act and the Stockholm Convention.
• This approach offers a valuable tool for prioritizing emerging organic contaminants.
• The findings highlight the need for proactive assessment of novel chemical substances.