Broadband Nanocavity Imaging with Machine Vision for Multiplex miRNA Assays.

Area of Science:

• Nanophotonics
• Biotechnology
• Artificial Intelligence

Background:

• Quantifying microRNAs (miRNAs) is difficult due to their low abundance and sequence similarity.
• Existing methods often require sample enrichment or amplification, limiting throughput and sensitivity.
• Complex biological matrices further complicate accurate miRNA detection.

Purpose of the Study:

• To develop a sensitive and multiplexed quantification method for microRNAs (miRNAs).
• To integrate a nanophotonic biosensor with deep learning for automated miRNA analysis.
• To demonstrate the platform's capability in analyzing endogenous miRNAs from lung cancer cell extracts.

Main Methods:

• Utilized a distributed Bragg reflector (DBR)-coupled silver nanoparticle (AgNP) gap nanocavity for enhanced signal collection and reduced quantum dot (QD) blinking.
• Employed deep learning instance segmentation (Mask R-CNN) for automated image analysis, counting, and classification of miRNAs.
• Applied the assay directly to A549 lung cancer cell extracts for quantifying endogenous miR-191, miR-25, and miR-130a without amplification.

Main Results:

• Achieved an ultra-low limit of detection (LOD) in the attomolar range (∼10^-17 mol/L).
• Demonstrated a broad linear dynamic range spanning five orders of magnitude.
• Obtained >99% correct identification accuracy across spectrally encoded channels for multiplexed analysis.

Conclusions:

• Established an AI-enabled nanophotonic biosensing platform for sensitive and specific miRNA detection.
• The platform offers a robust, scalable solution for multiplexed miRNA analysis in research and clinical settings.
• The developed assay directly quantifies endogenous miRNAs, overcoming limitations of traditional methods.