Adaptive Electrochemical Aptamer Sensing across Biofluid-Relevant pH Ranges via a Dual-Stage Machine Learning...

Area of Science:

• Electrochemistry
• Biomolecular Engineering
• Machine Learning

Background:

• Electrochemical sensing accuracy is often compromised by environmental factors like pH.
• Nonspecific adsorption and ionic strength variations further hinder reliable signal detection.
• Developing adaptive strategies is crucial for robust biosensing in variable conditions.

Purpose of the Study:

• To present a dual-stage machine learning framework for electrochemical sensing.
• To decouple environmental interference (pH) from analyte concentration-dependent signals.
• To enhance the quantitative accuracy of aptamer-based electrochemical sensors.

Main Methods:

• Utilized ferrocene-labeled DNA-gold nanoparticle (Fc-DNA@AuNP) probes.
• Employed machine learning classifiers to identify pH from electrochemical signal descriptors.
• Developed pH-specific regressors for concentration prediction.
• Validated performance across a pH range of 3.6-8.0.

Main Results:

• Machine learning classifiers achieved over 85% accuracy in identifying pH conditions.
• The framework successfully predicted carcinoembryonic antigen (CEA) concentrations over five orders of magnitude.
• Demonstrated adaptive interpretation of electrochemical signals without recalibration.

Conclusions:

• The developed framework provides a pH-adaptive interpretation strategy for aptamer-based electrochemical sensors.
• This approach enhances signal fidelity and quantitative accuracy in variable environments.
• Establishes a transferable signal-decoding layer for complex matrices and portable devices.