Deep Learning-Assisted Digital Microfluidic Platform for Automated CRISPR/Cas12 Detection of Mycobacterium...

Area of Science:

• Biotechnology
• Molecular Diagnostics
• Microfluidics

Background:

• CRISPR-based diagnostics offer high sensitivity and specificity but face challenges in point-of-care (POC) translation due to instrumentation and manual handling.
• Existing CRISPR diagnostic systems often require complex, costly equipment and extensive user intervention, limiting their accessibility.

Purpose of the Study:

• To develop a fully integrated, low-cost digital microfluidic (DMF) system for automating the complete CRISPR/Cas12 workflow.
• To enable sensitive and specific nucleic acid detection at the point-of-care, overcoming limitations of current benchtop systems.

Main Methods:

• Developed a DMF system with a programmable electrode array for droplet actuation and reagent mixing.
• Integrated a closed-loop heating module and a 3D-printed fluorescence imaging unit for automated sample processing and signal acquisition.
• Implemented a YOLOv11 deep learning model for objective, smartphone-based fluorescence signal interpretation.

Main Results:

• The DMF-CRISPR system successfully automated the entire CRISPR/Cas12 workflow, including sample preparation and detection.
• Achieved high analytical performance for Mycobacterium tuberculosis (MTB) DNA detection, with a dynamic range from 1 ng/μL down to 10⁻⁸ ng/μL.
• The deep learning model achieved a mean average precision of 0.889 for classifying results, demonstrating objective interpretation.

Conclusions:

• The integrated DMF-CRISPR system provides a portable, user-friendly, and cost-effective solution for point-of-care molecular diagnostics.
• The platform demonstrates comparable analytical performance to conventional methods while reducing user intervention and handling errors.
• This technology holds significant potential for accessible and rapid molecular diagnostics in resource-limited settings.