已核实 DOI 相关产品: 10

Integrating chromatography and mass spectrometry with multivariate machine learning for chemical warfare agents attribution

Journal 2026 · IF 5.2 · 发表于 2026 年 2 月 21 日
作者: Zixuan Zhang,Tengxin Huang,Xiaogang Lu,Runli Gao,Chengxin Pei,Junjie Ding,Hongmei Wang
阅读全文 在 PubMed 查看

摘要

Despite their prohibition under the Chemical Weapons Convention (CWC), chemical warfare agents (CWAs) remain in illicit use and pose a persistent threat to public safety and ecological integrity through accidental or intentional environmental release. Therefore, accurately identifying the source of these agents is critical for ensuring global chemical security. Herein, we develop a novel methodology that integrates comprehensive two-dimensional gas chromatography–time-of-flight mass spectrometry (GC × GC-TOFMS) with multiple machine learning techniques to discriminate the synthetic pathways of structurally analogous VX and VR nerve agents. A graph learning model was developed to transform chromatograms into discriminative features, enabling a subsequent Res-Net 50 classifier to achieve 94.4% and 100% accuracy in route discrimination for VX and VR, respectively. And a total of 163 and 143 chemical attribution signatures (CASs) were identified for VX and VR, respectively. Hierarchical cluster analysis (HCA) successfully discriminated the synthetic pathways with 100% accuracy for VX and 93.3% accuracy for VR, while both principal component analysis (PCA) and orthogonal partial least squares discriminant analysis (PLS-DA) achieved perfect classification with R 2  = 1. Notably, a self-organizing map (SOM) model was applied for the first time to chemical attribution. This study establishes a new paradigm for multi-machine-learning-assisted source tracing of highly similar environmental contaminants, enhancing the interpretation of complex chemical data and advancing the integration of analytical chemistry with data science for environmental forensics.

本研究中使用的阿拉丁产品

10 款产品在本文中被引用