Chen Dongyu, Wen Yumei, Li Ping, Zuo Can, Wang Yao, Wu Zhiyi
School of Automation and Intelligent Sensing, Shanghai Jiao Tong University, Dongchuan Road 800, Shanghai, 200240, China.
School of Automation and Intelligent Sensing, Shanghai Jiao Tong University, Dongchuan Road 800, Shanghai, 200240, China.
Talanta. 2026 Jan 1;296:128532. doi: 10.1016/j.talanta.2025.128532. Epub 2025 Jun 30.
Accurate and synchronized assessment of biochemical parameters, such as biomarker concentration and body fluid viscosity, is crucial for advancing early disease detection and health management. Conventional biomolecular multiparameter detection methods often rely on multiple sensors or analytical techniques, which introduce cross-talk between sensing modalities, data inconsistencies, and complex calibration requirements, ultimately compromising detection precision and adaptability. We propose a streamlined detection approach that leverages a single uncoated Quartz Crystal Microbalance (QCM) sensor to monitor the dynamic magnetized motion of biomolecules under multimodal magnetic field modulation. Unlike conventional QCM methods that rely on static mass loading effects, this approach enables the sensor to capture motion signals that encode information about biomolecule concentration and base liquid viscosity. A backpropagation (BP) neural network is employed to model the nonlinear coupling between these motion-derived signal characteristics and the target biochemical parameters. The proposed method is validated using prostate-specific antigen (PSA) as a biomolecular model analyte. Experimental results from blind tests, where both concentration and viscosity were simultaneously unknown, demonstrate a prediction accuracy of 90 % for concentrations ranging from 0.01 to 1000 ng/mL and 87 % for viscosities between 1 and 6 cP. By integrating multimodal magnetic modulation with QCM-based motion sensing and machine learning, the BP-MMM-QCM technique provides a versatile and high-precision solution for biomolecule analysis. Accurate detection of biomolecule concentrations is essential for early disease diagnosis as well as monitoring disease progression and therapeutic responses. This approach overcomes the limitations of conventional QCM methods and enables real-time, multi-parameter detection in a single assay, making it a promising tool for disease diagnostics and health monitoring applications.
准确且同步地评估生物化学参数，如生物标志物浓度和体液粘度，对于推进疾病早期检测和健康管理至关重要。传统的生物分子多参数检测方法通常依赖于多个传感器或分析技术，这会引入传感模式之间的串扰、数据不一致以及复杂的校准要求，最终影响检测精度和适应性。我们提出了一种简化的检测方法，该方法利用单个未涂层的石英晶体微天平（QCM）传感器，在多模态磁场调制下监测生物分子的动态磁化运动。与依赖静态质量负载效应的传统QCM方法不同，这种方法使传感器能够捕获编码生物分子浓度和基础液体粘度信息的运动信号。采用反向传播（BP）神经网络对这些运动衍生信号特征与目标生物化学参数之间的非线性耦合进行建模。以前列腺特异性抗原（PSA）作为生物分子模型分析物对所提出的方法进行了验证。在浓度和粘度均未知的盲测实验结果表明，对于浓度范围为0.01至1000 ng/mL的情况，预测准确率为90%，对于粘度在1至6 cP之间的情况，预测准确率为87%。通过将多模态磁调制与基于QCM的运动传感和机器学习相结合，BP-MMM-QCM技术为生物分子分析提供了一种通用且高精度的解决方案。准确检测生物分子浓度对于早期疾病诊断以及监测疾病进展和治疗反应至关重要。这种方法克服了传统QCM方法的局限性，能够在单次检测中进行实时多参数检测，使其成为疾病诊断和健康监测应用的有前途的工具。
