Xu Shiyuan, Huang Yi, Ma Dannv, Wu Jiaying, Liu Xuemei, Zhang Qianru, Gu Zhuangyue, Pan Aiwu, Wu Jianmin
Lab of Nanomedicine and Omic-Based Diagnostics, Institute of Analytical Chemistry, Department of Chemistry, Zhejiang University, Hangzhou 310058, P. R. China.
Department of Internal Medicine, The Second Affiliated Hospital of Zhejiang University School of Medicine, Hangzhou 310003, China.
ACS Nano. 2025 May 27;19(20):19429-19441. doi: 10.1021/acsnano.5c04231. Epub 2025 May 16.
As a prevalent clinical condition, it is critical to distinguish between bacterial and viral respiratory tract infections given their pivotal role in guiding appropriate pharmaceutical interventions and preventing antibiotic misuse. Exhaled breath (EB) contains a spectrum of disease-specific biomarkers, enabling precise diagnostic analysis. Thus, EB analysis using an electronic nose (e-nose) to record electrical response fingerprints and discriminate pathogens via machine learning algorithms has emerged as a promising noninvasive diagnostic technology. In this study, a graphene-based e-nose sensor array modified with metal-organic frameworks (MOFs) and metal phthalocyanines (MPcs) was developed by using multiple reduction methods. The sensor array demonstrated excellent capability in distinguishing between two types of EB samples collected from healthy individuals spiked with acetone and isoprene, which are closely associated with bacterial and viral respiratory infections. Furthermore, a diagnostic model was constructed using e-nose fingerprints from 145 clinical EB samples comprising 89 bacterial infection cases and 56 viral infection cases. A weighted fusion classification model, integrating the support vector machine, random forest, and Lasso regression (Lasso), achieved an accuracy of 83.7% in the validation group, with an area under the curve (AUC) of 0.87. An independent external clinical trial involving 43 respiratory infection patients (including 6 unidentified cases) yielded an accuracy of 75.7% and an AUC of 0.81 for distinguishing bacterial from viral infections. Additionally, the sensor array achieved a 75% accuracy rate in discriminating mycoplasma infections by using linear discriminant analysis. These results suggest that the graphene-based e-nose array modified with MOFs and MPcs is a promising tool for diagnosing respiratory tract infections, aiding in optimized treatment decisions and potentially improving therapeutic efficiency.
作为一种常见的临床病症，区分细菌和病毒引起的呼吸道感染至关重要，因为这对于指导恰当的药物干预以及防止抗生素滥用起着关键作用。呼出气体（EB）包含一系列疾病特异性生物标志物，能够实现精确的诊断分析。因此，利用电子鼻（e-nose）记录电响应指纹并通过机器学习算法辨别病原体的EB分析已成为一种有前景的非侵入性诊断技术。在本研究中，通过多种还原方法开发了一种用金属有机框架（MOF）和金属酞菁（MPc）修饰的基于石墨烯的电子鼻传感器阵列。该传感器阵列在区分从添加了丙酮和异戊二烯的健康个体采集的两类EB样本方面表现出卓越能力，丙酮和异戊二烯与细菌和病毒呼吸道感染密切相关。此外，利用来自145份临床EB样本（包括89例细菌感染病例和56例病毒感染病例）的电子鼻指纹构建了一个诊断模型。一个整合了支持向量机、随机森林和套索回归（Lasso）的加权融合分类模型在验证组中准确率达到83.7%，曲线下面积（AUC）为0.87。一项涉及43例呼吸道感染患者（包括6例未明确诊断病例）的独立外部临床试验在区分细菌感染和病毒感染方面准确率为75.7%，AUC为0.81。此外，该传感器阵列通过线性判别分析在区分支原体感染方面准确率达到75%。这些结果表明，用MOF和MPc修饰的基于石墨烯的电子鼻阵列是诊断呼吸道感染的一种有前景的工具，有助于优化治疗决策并可能提高治疗效率。
