Machine learning-powered fluorescent sensor arrays for rapid detection of heavy metals and pesticides in complex environments

Ming-Ming Chen; Yan-Qing Zhang; Lu-Chen Cheng; Fang-Jie Zhao; Peng Wang

doi:10.1016/j.bios.2025.117706

Machine learning-powered fluorescent sensor arrays for rapid detection of heavy metals and pesticides in complex environments

Biosens Bioelectron. 2025 Jun 18:287:117706. doi: 10.1016/j.bios.2025.117706. Online ahead of print.

Authors

Ming-Ming Chen¹, Yan-Qing Zhang², Lu-Chen Cheng², Fang-Jie Zhao², Peng Wang³

Affiliations

¹ Centre for Agricultural and Environmental Health, Academy for Advanced Interdisciplinary Studies, Nanjing Agricultural University, Nanjing, 210095, China.
² Jiangsu Collaborative Innovation Center for Solid Organic Waste Resource Utilization, College of Resources and Environmental Sciences, Nanjing Agricultural University, Nanjing, 210095, China.
³ Centre for Agricultural and Environmental Health, Academy for Advanced Interdisciplinary Studies, Nanjing Agricultural University, Nanjing, 210095, China; Jiangsu Collaborative Innovation Center for Solid Organic Waste Resource Utilization, College of Resources and Environmental Sciences, Nanjing Agricultural University, Nanjing, 210095, China. Electronic address: p.wang3@njau.edu.cn.

PMID: 40540972
DOI: 10.1016/j.bios.2025.117706

Abstract

The co-contamination of multiple pollutants in complex environmental matrices poses a significant threat to ecosystems and public health, necessitating advanced detection methods. In this study, we developed a machine learning-powered chemical sensor array capable of simultaneously identifying and discriminating nine heavy metal(loid)s (Cr[III], Cd[II], Hg[II], Pb[II], Co[II], Zn[II], Mn[II], As[III], and Se[VI]) and five pesticides (propiconazole, penconazole, cyproconazole, indoxacarb, and azoxystrobin). Using three distinct copper nanoclusters (Cu NCs) with unique ligand-based binding affinities, the system generated characteristic fluorescent "fingerprints". By coupling with machine-learning algorithms (LDA and HCA), the sensor array achieved 100 % identification accuracy within 10 min, with exceptional sensitivity (limits of detection: ∼0.5 nM for heavy metal(loid)s and ∼7.1 ppb for pesticides). This approach was validated using real-world samples, including blood, urine, soil, tap water, vegetables, and fruits, demonstrating high selectivity, anti-interference capability, and practical applicability. This proposed nanosensor array provides a robust, rapid, and sensitive platform for multi-target detection, offering transformative solutions in food safety, environmental monitoring, and public health surveillance.

Keywords: Copper nanoclusters; Florescent sensor array; Heavy metals detection; Machine learning; Pesticides identification.