FluoroMatch 2.0-making automated and comprehensive non-targeted PFAS annotation a reality

Abstract

Because of the pervasiveness, persistence, and toxicity of per- and polyfluoroalkyl substances (PFAS), there is growing concern over PFAS contamination, exposures, and health effects. The diversity of potential PFAS is astounding, with nearly 10,000 PFAS catalogued in databases to date (and growing). The ability to detect the thousands of known PFAS, and discover previously uncatalogued PFAS, is necessary to understand the scope of PFAS contamination and to identify appropriate remediation and regulatory solutions. Current non-targeted methods for PFAS analysis require manual curation and are time-consuming, prone to error, and not comprehensive. FluoroMatch Flow 2.0 is the first software to cover all steps of data processing for PFAS discovery in liquid chromatography-high-resolution tandem mass spectrometry samples. These steps include feature detection, feature blank filtering, exact mass matching to catalogued PFAS, mass defect filtering, homologous series detection, retention time pattern analysis, class-based MS/MS screening, fragment screening, and predicted MS/MS from SMILES structures. In addition, a comprehensive confidence level criterion is implemented to help users understand annotation certainty and integrate various layers of evidence to reduce overreporting. Applying the software to aqueous film forming foam analysis, we discovered over one thousand likely PFAS including previously unreported species. Furthermore, we were able to filter out 96% of features which were likely not PFAS. FluoroMatch Flow 2 increased coverage of likely PFAS by over tenfold compared to the previous release. This software will enable researchers to better characterize PFAS in the environment and in biological systems.

Keywords: Aqueous film forming foam; Liquid chromatography; Mass spectrometry; Non-targeted analysis; PFAS; Software.

Fig. 1

Evidence used for annotation in FluoroMatch 2.0 (retention time (RT), exact mass from full-scan data (MS), and fragmentation (MS/MS))

Fig. 2

Two different methods for generating MS/MS libraries of PFAS (i and iii) and three different methods for MS/MS annotation are shown (ii and iv). PFAS MS/MS rules are assigned for each PFAS class with standards acquired or annotated spectra from the literature (i), which are used for confident annotations (ii). Furthermore, tentative confident annotations with one or more of these class-based fragments observed are also assigned (ii). “In silico” fragmentation libraries which are more comprehensive in terms of fragment coverage but potentially less accurate are developed by assigning substructure-fragment relationships (iii). In this case, fragmentation is predicted from molecules in the EPA structure list. Tentative annotations using this library are made if F-containing fragments are observed (iv)

Fig. 3

Scoring system implemented in FluoroMatch 2.0. Acronyms: MS—in this case refers to exact mass matches; MS/MS—refers to any algorithms employed by FluoroMatch 2.0 using exact mass to match experimental fragmentation to known PFAS fragments; RT—retention time; homologous series—in this case PFAS in the same class but with a different number of CF₂ repeating units determined by exact mass intervals (50 for CF₂) and the same CF₂-normalized mass defect

Fig. 4

FluoroMatch 2.0 user workflow and data processing steps covered by the fully automated version (FluoroMatch Flow) and the modular version (FluoroMatch Modular). Acronyms: DDA, data-dependent analysis; IE, iterative exclusion; LC-HRMS, liquid chromatography–high-resolution mass spectrometry; MS/MS, tandem mass spectrometry. Tan and gray boxes indicate aspects of both automated and modular workflows

Fig. 5

Overview of features proposed as potential PFAS, their scores, fragmentation evidence, and formation of homologous series (see Fig. 3 for definitions of scores). i The number and percent of features under each general scoring criteria and broad definition of scoring criteria (e.g., A is the sum of A+, A, and A−). ii Breakdown of the number of features across more refined scoring criteria A+, A, and A− (“−” for A and B mean the compounds were not observed within homologous series). iii Homologous series with 3 or more members; the top score for each series was used for classification by general scoring criteria for this figure (color codes are the same as in Fig. 5i). iv Number of features with new MS/MS assignments (score A or B) during iterative injections with a rolling exclusion list (color codes are the same as in Fig. 5i). v Number of fluorine-containing fragments using fragment screening (777 fragments screened) for features, categorized by score

Fig. 6

Examples of PFAS homologous series automatically determined using FluoroMatch 2.0 with confident annotations (at least one score A: colored shapes), select tentative annotations (at least one score B: gray shapes), and newly discovered PFAS (yellow square with black lines). Red outlines are compounds flagged as tentative, for example, due to retention times not being in the correct order

Fig. 7

Identification of unknown-unknowns (PFSA—pentafluorosulfide mono-unsaturated) two–three members detected using fragment screening. Comparison of a validated fully saturated species is shown for reference