Artificial neural network modeling and optimization of an electrochemical biosensor for plasma miR-155-based breast cancer detection

Abstract

MicroRNA-155 (miR-155) is a clinically important biomarker involved in cancer progression, immune regulation, and inflammatory diseases, highlighting the need for sensitive and reliable detection methods. Conventional biosensor fabrication often relies on labor-intensive trial-and-error optimization, which delays the development of practical diagnostic tools. In contrast to most previous studies that focus on predicting analyte concentration from biosensor signals, this work develops a data-driven framework for modeling the nonlinear relationships between fabrication parameters and biosensor output. Artificial neural networks (ANN) and adaptive neuro-fuzzy inference systems (ANFIS) were proposed to model a voltammetric biosensor for plasma miR-155 detection. A dataset containing the biosensor output current and six fabrication parameters was used. The optimal parameter values were determined using a genetic algorithm (GA). The results show that the ANN approach outperforms ANFIS, achieving an value of 0.9845. The optimal fabrication parameters were 7.12 nM, 85.22 min, 6.54 min, 118.02 min, 0.12 mM, and 93.39 min for detection probe concentration, detection probe incubation time, MCH incubation time, hybridization time, OB concentration, and OB incubation time, respectively, resulting in an output current of 223 nA. The ANN-GA framework offers a practical and efficient strategy for biosensor development by reducing experimental iterations, thereby lowering material consumption and enabling rapid parameter optimization. These findings demonstrate that ANN-assisted optimization can accelerate the development of cost-effective, high-performance biosensors, supporting their translation into clinical diagnostics for early and accurate miR-155 detection.