Engineering an integrated biosensing interface combining DNA-assisted clustering and explainable AI for biomarker...

Area of Science:

• Biomedical Engineering
• Biosensing Technology
• Artificial Intelligence in Diagnostics

Background:

• Point-of-care testing (POCT) faces challenges in reliable signal readout within complex clinical samples.
• Existing molecular amplification methods often struggle with consistency in real-world biological matrices.

Purpose of the Study:

• To develop an integrated biosensing framework prioritizing readout reliability as a core engineering goal.
• To create a generalizable strategy for trustworthy POCT systems by co-designing sensing and interpretation.

Main Methods:

• Utilized a heptameric nanobody probe for multivalent target capture.
• Employed a DNA-assisted clustering interface for gold nanoparticle reporter amplification.
• Integrated a few-shot learning module (Prototypical Networks) for robust, interpretable classification.

Main Results:

• Achieved a visual limit of detection of 2 ng/mL for alpha-fetoprotein.
• Demonstrated quantitative consistency across diverse clinical serum samples.
• The AI module effectively mitigated matrix effects and imaging variability for reliable readout.

Conclusions:

• The integrated framework successfully addresses POCT readout reliability challenges.
• This co-design approach of sensing and interpretation offers a generalizable strategy for trustworthy diagnostic platforms.
• The system shows promise for clinical applications like hepatocellular carcinoma screening.