Deep learning enhancing guide RNA design for CRISPR/Cas12a-based diagnostics

Area of Science:

• Molecular Biology
• Bioinformatics
• Infectious Disease Diagnostics

Background:

• Rapid and accurate diagnostic tests are crucial for patient outcomes and controlling infectious diseases.
• CRISPR Cas12a systems offer promising on-site nucleic acid testing capabilities.
• Designing effective CRISPR RNA (crRNA) for Cas12a detection is currently challenging and time-consuming.

Purpose of the Study:

• To develop an enhanced, AI-driven system for designing crRNA for Cas12a-based diagnostics.
• To improve the speed and accuracy of crRNA design for Cas12a detection systems.
• To create a user-friendly platform integrating crRNA and primer design for enhanced accessibility.

Main Methods:

• Developed EasyDesign, a deep learning system utilizing an optimized convolutional neural network (CNN).
• Trained the CNN model on 11,496 experimentally validated Cas12a detection cases across various pathogens.
• Assessed model performance on novel pathogens and integrated it into an interactive web server with RPA primer design.

Main Results:

• Achieved a high prediction accuracy (Spearman's ρ = 0.812) for crRNA design.
• Demonstrated superior prediction performance compared to traditional experimental screening for new pathogens.
• Successfully designed functional crRNAs for HPV subtypes, showing robust fluorescent signals in assays.

Conclusions:

• EasyDesign provides a rapid, reliable, and accurate solution for crRNA design in Cas12a diagnostics.
• The integrated web server enhances accessibility for researchers and clinicians.
• This platform has the potential to significantly facilitate clinical sample testing and research applications.