Document Type : Research Paper
microRNAs are a class of small non-coding RNA molecules that play a central role in post-transcriptional gene regulation and are critically involved in plant growth, development, and responses to biotic and abiotic stresses. Under biotic stress conditions such as pathogen infection, miRNAs rapidly respond by fine-tuning the expression of defense-related genes, making them promising molecular indicators for early stress detection in plants. However, the extremely low endogenous concentration of plant miRNAs, often in the femtomolar or attomolar range, poses major challenges for their reliable quantification using conventional techniques such as qPCR, microarrays, or northern blotting. In recent years, electrochemical biosensors have emerged as powerful analytical tools for sensitive miRNA detection due to their simplicity, high sensitivity, and potential for miniaturization. Nevertheless, the performance of such biosensors strongly depends on the optimal configuration of key design parameters, including probe density and nanomaterial loading. Therefore, this study aimed to develop and optimize an electrochemical miRNA biosensor using a machine learning–based approach and to evaluate its applicability for detecting brown spot disease–induced biotic stress in barley plants through quantification of miRNA168 and miRNA397.
An electrochemical microRNA biosensor was constructed based on the hybridization of target microRNAs with thiolated complementary oligonucleotide probes immobilized on a gold working electrode. Following hybridization, positively charged polyethyleneimine–silver nanoparticles (PEI–Ag NPs) were electrostatically attracted to the negatively charged probe–microRNA duplex, generating an electrochemical signal proportional to the target microRNA concentration. Cyclic voltammetry measurements were performed using a three-electrode system. To optimize biosensor performance, two key parameters—the probe volume (5–10 µL) and the mass of PEI–Ag NPs (1–6 µg)—were varied, resulting in a dataset of 100 experimental observations obtained at a target microRNA concentration of 1 fM. A support vector regression model with a radial basis function kernel was developed to model the nonlinear relationship between input variables and the biosensor response. Model hyperparameters were optimized using a genetic algorithm, and five-fold cross-validation was applied to assess predictive performance and prevent overfitting. After model validation, a dense input grid was generated to identify the global optimum configuration. The optimized biosensor was then calibrated using various dilutions of standard microRNA. Barley plants (Hordeum vulgare L., susceptible cultivar ‘Sahra’) were grown under controlled greenhouse conditions and inoculated with the spot blotch pathogen Bipolaris sorokiniana. Fourteen days after inoculation, total RNA was extracted from leaf tissues, and microRNA concentrations were measured using the optimized biosensor.
The support vector regression model demonstrated excellent predictive accuracy, with a mean squared error of 0.004 and a coefficient of determination (R²) of 0.97 across cross-validation folds, indicating strong generalization capability. The optimal biosensor configuration was identified at a probe volume of 9.15 mL and a PEI–Ag NP mass of 1.85 µg, neither of which lay at the boundaries of the tested ranges, confirming the adequacy of the experimental design space. The optimized biosensor exhibited a wide linear detection range from 10⁻¹⁹ to 10⁻¹¹ M, along with good repeatability. Application of the biosensor to plant samples revealed significant alterations in miRNA168 and miRNA397 concentrations in infected barley leaves compared with healthy controls. The observed increase in miRNA168 is consistent with its regulatory role in targeting AGO1, suggesting a feedback mechanism to balance RNA silencing activity during immune activation. Similarly, elevated miRNA397 levels imply modulation of laccase-mediated lignification pathways, potentially favoring the accumulation of antimicrobial phenolic compounds during early infection stages. These findings demonstrate that the developed biosensor is capable of capturing biologically meaningful microRNA expression changes associated with pathogen-induced stress.
This study presents a machine learning–assisted strategy for optimizing an electrochemical microRNA biosensor and demonstrates its successful application for detecting biotic stress in barley plants. The integration of support vector regression with experimental biosensor data enabled accurate modeling and efficient identification of optimal design parameters, resulting in high sensitivity and a broad linear detection range. The optimized biosensor effectively discriminated between healthy and diseased plants based on miRNA expression profiles, highlighting its potential as a rapid and sensitive tool for early plant stress diagnostics. Although environmental variability and the multifactorial regulation of miRNA expression remain challenges for large-scale field deployment, the proposed approach provides a strong foundation for developing advanced biosensing platforms for precision agriculture and plant health monitoring.
The study was funded by the Gorgan University of Agricultural Sciences and Natural Resources under the title of Approved Grants for PhD Dissertations.
K.R. and K.A.: Conceptualization; N.D.: Methodology; N.D. and K.A.: Formal analysis; N.D.: Writing—original draft preparation; K.R. and K.A.: Writing—review and editing. All authors have read and agreed to the published version of the manuscript.
Data is available upon reasonable request from the corresponding author.
The authors would like to thank the Gorgan University of Agricultural Sciences and Natural Resources for funding.
The authors avoided data fabrication, falsification, and plagiarism, and any form of misconduct.
The authors declare no conflict of interest.
