Integration of Per- and Polyfluoroalkyl Substance (PFAS) Fingerprints in Fish with Machine Learning for PFAS Source Tracking in Surface Water

Per- and polyfluoroalkyl substances (PFASs) are a classof environmentalcontaminants that originate from various sources. The unique chemicalfingerprints associated with many commercial products and industrialapplications make PFASs ideal candidates for machine learning (ML)-assistedenvironmental forensics. Here, we propose a novel use of PFAS fingerprintsin fish tissue from surface water systems to classify exposure frommultiple sources of PFASs using a proof-of-concept demonstration.Three supervised ML classification techniques (k-nearest neighbors(KNN), decision trees, support vector machines) implementing two predictivefeatures are used to classify literature-reported PFAS fingerprintsin fish (n = 1057). The importance of additionalpredictive features was explored using brute force optimization ofa multifeature KNN algorithm. The multiclass classification consideredexposure to aqueous film-forming foam-impacted water, paper industrywastewater, diffuse sources, or PFASs undergoing long-range transport.The optimized classifiers demonstrated 85%-94% classificationaccuracy for this first known multiclass classification of PFASs forenvironmental forensics. The optimized classifiers also demonstrated79%-92% classification accuracy with a set of independent externalvalidation data (n = 192). Our results demonstratethat PFAS fingerprints in fish tissue may be an effective means ofPFAS source tracking in surface water systems. The source code isprovided for guidance on best practices for ML-assisted environmentalforensics.