Multidimensional feature reinforced single-molecule identification of aromatic polyfluorinated carboxylic acid isomers

Per- and polyfluorinated carboxylic acids (PFCAs) are a large class of persistent, bioaccumulative substances that pose well-established threats to ecosystems and human health. Recent advances in the single-molecule electrochemical sensing technology based on nanopore electrochemistry has made significant progress towards standard-free detection of PFCAs by establishing a strict linear correlation between the current blockade of PFCAs and their volume. However, the spatial resolution of current approaches remains insufficient to identify PFCAs isomers with subtle structural and volumetric differences. In this study, we introduced an integrated strategy combining single-molecule electrochemical sensing with machine learning to identify 17 aromatic PFCAs isomers across three categories. By extracting multidimensional features from raw single-molecule signals and applying low-pass filtering for noise reduction, we effectively captured intrinsic molecular information embedded within the signals, achieving an overall classification accuracy of 88.92%. Furthermore, anti-interference test confirmed the potential for practical application in complex environmental samples. This work lays a solid foundation for achieving comprehensive, label-free detection of PFCAs.

全氟和多氟羧酸（Per- and polyfluorinated carboxylic acids, PFCAs）是一类持久性、生物累积性物质，已被证实对生态系统和人类健康构成明确威胁。近年来，基于纳米孔电化学的单分子电化学传感技术取得重要进展，通过建立PFCAs的电流阻塞效应与其分子体积间的严格线性相关性，在PFCAs的无标检测领域实现了突破。然而，现有方法的空间分辨率仍不足以区分结构与体积差异细微的PFCAs异构体。本研究提出了一种融合单分子电化学传感与机器学习的整合策略，用于识别三类共17种芳香族PFCAs异构体。通过从原始单分子信号中提取多维特征并采用低通滤波进行降噪，我们有效捕获了信号中蕴含的内在分子信息，整体分类准确率达88.92%。此外，抗干扰测试证实了该方法在复杂环境样品中实际应用的潜力。本研究为实现PFCAs的全面无标记检测奠定了坚实基础。