FMS2Lib: A Machine Learning-Ready Tandem Mass Spectral Library for Nontargeted Analysis of Non-conventional PFAS
Abstract
Nontargeted identification of non-conventional per- and polyfluoroalkyl substances (PFAS), including fluorinated pharmaceuticals and pesticides, remains challenging. Existing machine learning models for nontargeted analysis, primarily trained on tandem mass spectra of nonfluorinated compounds, may not generalize well to non-conventional PFAS because of their distinct mass spectrometric chemistry. Here, we established FMS2Lib, an experimental tandem mass spectral library containing 34,686 fluorinated drug-like compounds (F 1-4 ), to support machine learning-based nontargeted PFAS analysis. We demonstrated that training with FMS2Lib substantially improved the performance of existing models for both molecular formula prediction and structure annotation. We further applied these models to identify PFAS in passive samplers from wastewater effluents and surface waters and in high-throughput laboratory biotransformation experiments, revealing dozens of non-conventional PFAS and their biotransformation products. Collectively, FMS2Lib addresses a major data bottleneck in machine learning-based nontargeted analysis of non-conventional PFAS.
Related
- Beyond conventional sequence-based encoding: a generalized Machine Learning pipeline for oligonucleotide property prediction
- An integrated machine learning framework for accurate, real-time and non-destructive vitamin C quantification from Vis/NIR spectral data
- Comparing Suspect and Nontargeted Screening Workflows for Perinatal Chemical Exposome Analysis: A European Interlaboratory Study from the PARC-EU Project
- Predicting fluorinated gas solubility in deep eutectic solvents using machine learning and deep learning approaches
- Machine Learning for NMR Observables Sensitive to Nuclear Quantum Effects
BibTeX
@misc{yang2026fms2lib,
title = {{FMS2Lib: A Machine Learning-Ready Tandem Mass Spectral Library for Nontargeted Analysis of Non-conventional PFAS}},
author = {Tao Yang and Xiaoyun Wang and Diwen Yang and Elizabeth Xu and Lifei Wang and Sonya M. Kleywegt and Paul Helm and Frank Wania and Xianzhi Peng and Hui Peng},
howpublished = {ChemRxiv},
year = {2026},
month = oct,
doi = {10.26434/chemrxiv.15010212/v1},
url = {https://chemrxiv.org/doi/full/10.26434/chemrxiv.15010212/v1},
note = {CC BY},
}
RIS
TY - UNPB
TI - FMS2Lib: A Machine Learning-Ready Tandem Mass Spectral Library for Nontargeted Analysis of Non-conventional PFAS
AU - Tao Yang
AU - Xiaoyun Wang
AU - Diwen Yang
AU - Elizabeth Xu
AU - Lifei Wang
AU - Sonya M. Kleywegt
AU - Paul Helm
AU - Frank Wania
AU - Xianzhi Peng
AU - Hui Peng
PY - 2026
DA - 2026/10/11
PB - ChemRxiv
DO - 10.26434/chemrxiv.15010212/v1
UR - https://chemrxiv.org/doi/full/10.26434/chemrxiv.15010212/v1
AB - Nontargeted identification of non-conventional per- and polyfluoroalkyl substances (PFAS), including fluorinated pharmaceuticals and pesticides, remains challenging. Existing machine learning models for nontargeted analysis, primarily trained on tandem mass spectra of nonfluorinated compounds, may not generalize well to non-conventional PFAS because of their distinct mass spectrometric chemistry. Here, we established FMS2Lib, an experimental tandem mass spectral library containing 34,686 fluorinated drug-like compounds (F 1-4 ), to support machine learning-based nontargeted PFAS analysis. We demonstrated that training with FMS2Lib substantially improved the performance of existing models for both molecular formula prediction and structure annotation. We further applied these models to identify PFAS in passive samplers from wastewater effluents and surface waters and in high-throughput laboratory biotransformation experiments, revealing dozens of non-conventional PFAS and their biotransformation products. Collectively, FMS2Lib addresses a major data bottleneck in machine learning-based nontargeted analysis of non-conventional PFAS.
ER -