Suspect and nontarget screening of per- and polyfluoroalkyl substances based on ion mobility mass spectrometry and machine learning techniques.

High-resolution mass spectrometry (HRMS)-based suspect and nontarget screening techniques are powerful tools for the comprehensive identification of per- and polyfluoroalkyl substances (PFASs), but the interference of complex matrices (especially for wastewater) pose an analytical challenge. This study explored the potential of combining ion mobility spectrometry (IMS) with HRMS and machine learning techniques to achieve the rapid and accurate suspect and nontarget screening of PFAS in wastewater. There were fewer interfering peaks and a clearer spectrum in the data acquired by IMS-HRMS than conventional HRMS. The introduction of collision cross section (CCS) in PFAS homologous series search could filter out 63% of false positive results. Retention time and CCS prediction models were helpful in improving the confidence for PFAS qualitative identification and the random forest algorithm combined with RDKit descriptor performed best for CCS prediction. With the inclusion of extra dimensional information, this study also proposed a comprehensive and concise confidence assignment criterion to better convey the certainty of the qualitative identification of PFAS. Finally, a total of 56 potential PFASs were identified in the wastewater sample using the newly developed method and 45 of them were identified outside reference standards, emphasizing the importance of suspect and nontarget screening for PFAS.