Cascaded SDA-Exo III amplification with machine learning-driven multi-modal data fusion for self-powered biosensing.

BACKGROUND

Pokkah boeng disease, caused by Fusarium spp., is a destructive disease that can leads to substantial sugarcane yield losses under epidemic conditions. Early detection is extremely challenging due to the absence of visible symptoms during early-stage infection. Conventional diagnostic methods rely on sophisticated instrumentation and external power sources, which fundamentally restrict their practical application in resource-limited agricultural settings. To address these limitations, we develop a smart biosensing platform that breaks through the constraints of traditional methods, enabling effective disease management and sugarcane cultivation in resource-limited environments.

RESULTS

We develop a smart biosensing platform that integrates biomolecular circuits, triple-modal signal transduction, and machine learning for automated standard curve construction. The core innovation lies in a cascaded nucleic acid amplification strategy, where strand displacement amplification is combined with Exo III-mediated cleavage to physically separate the signals, enabling the output of electrochemical, colorimetric, and photothermal signals. This design provides built-in cross-validation, reducing false positives and false negatives in complex matrices. The platform utilizes functional nanomaterials to enhance performance: ZIF-67/AuNPs improve electron transfer at the bioanode of a self-powered biofuel cell, while TiO2@Pt nanoparticles ensure efficient photothermal conversion. Machine learning algorithms (Linear Regression, Ridge Regression, Lasso Regression, and Elastic Net) intelligently fuse multi-source data from the three detection modalities. The Elastic Net model achieves optimal performance, with feature importance analysis indicating that the photothermal mode contributes most significantly, followed by electrochemical and colorimetric signals. The platform attains excellent limits of detection: 16.92 aM for the electrochemical mode, 21.75 aM for the colorimetric mode, and 2.86 fM for the photothermal mode, alongside high stability and reproducibility.

SIGNIFICANCE: This work integrates molecular biology, nanomaterials, and data science, bridging the entire chain from the construction of a biological detection platform to data output and intelligent data analysis, and establishes a new paradigm for intelligent biosensors capable of automatically constructing standard curves in precision agriculture.