文献类型：期刊文献

英文题名：High Na-Adducted Ion Selectivity in Laser Desorption/Ionization Mass Spectrometry with a Steric-Tuned COF

作者：Yan, Zhichao[1];Wang, Xiaoyan[1];Wang, Zhenxin[2];Zheng, Yanfeng[1];Zhang, Lingyi[1];Zhang, Weibing[1]

机构：[1]East China Univ Sci & Technol, Sch Chem & Mol Engn, Shanghai Key Lab Funct Mat Chem, Shanghai 200237, Peoples R China;[2]Fudan Univ, Zhongshan Hosp, Dept Lab Med, Shanghai 200032, Peoples R China

年份：2026

卷号：98

期号：1

起止页码：410

外文期刊名：ANALYTICAL CHEMISTRY

收录：;EI(收录号：20260319906098);WOS:【SCI-EXPANDED(收录号:WOS:001648546100001)】；

基金：We gratefully acknowledged the support of the Fundamental Research Funds for the Central Universities (JKJ01251717).

语种：英文

外文关键词：Amino acids - Binding energy - Binding sites - Biomolecules - Design for testability - Desorption - Diagnosis - Drug products - Laser diagnostics - Learning algorithms - Learning systems - Metabolism - Metabolites - Substrates - Support vector machines - Viruses

摘要：Laser desorption-ionization mass spectrometry (LDI-MS) is prized for its rapidity, simplicity, low sample consumption, and high throughput in metabolic analysis, with great potential as a diagnostic tool. Herein, an LDI-MS method with sulfone-containing covalent organic frameworks (S-COFs) as the substrate was developed. By engineering the electron cloud density and steric hindrance of the LDI substrate, selective binding sites with higher affinity for Na+ than K+ were obtained. DFT calculations confirmed that the sulfone group (O-O distance: 2.14 angstrom; O=S=O angle: 117.43 degrees) favors Na+ adsorption over K+. For glucose, the intensity of [M + Na](+) was 75-fold higher than [M+K](+), enhancing detection sensitivity while simplifying serum metabolic fingerprints. S-COF-assisted LDI-MS exhibited robust responses to carbohydrates, amino acids, nucleosides, and other small-molecule metabolites. It achieved an excellent limit of detection (LOD) of 10 pg for glucose, 1,926 times more sensitive than the conventional organic matrix DHB. The method was successfully applied to detect serum metabolites in hepatitis C virus (HCV) patients and healthy controls. Support vector machine (SVM) and neural network (NNW) machine learning algorithms enabled high-precision HCV diagnosis, both yielding an AUC of 0.990. 36 metabolites were identified, and 5 of them were strongly enriched in the cysteine/methionine pathway (-log P = 5.70, Impact = 0.347), implicating perturbed redox homeostasis. This study not only offers novel insights into the modular design of functionalized COFs as LDI substrates but also provides a versatile tool for large-scale clinical diagnostics.