Classification and specific primer design for accurate detection of SARS-CoV-2 using deep learning

Area of Science:

• Genomics
• Artificial Intelligence
• Virology

Background:

• Accurate and rapid detection of SARS-CoV-2 is critical for pandemic control.
• Existing diagnostic methods may require optimization for speed and specificity.
• Identifying unique viral genomic sequences aids in developing targeted diagnostic tools.

Purpose of the Study:

• To develop a novel methodology for discovering representative genomic sequences of SARS-CoV-2 using deep learning and explainable AI.
• To validate the efficacy of discovered sequences in differentiating SARS-CoV-2 from other virus strains.
• To design and test a diagnostic primer set based on the identified sequences.

Main Methods:

• A convolutional neural network classifier was trained on SARS-CoV-2 genomic sequences.
• Explainable AI techniques were employed to analyze the network's decision-making process.
• Discovered sequences were validated using data from public repositories and patient samples.
• A primer set was designed from a representative sequence and tested against existing methods.

Main Results:

• The convolutional neural network achieved 98.73% accuracy in classifying virus strains.
• AI analysis identified exclusive SARS-CoV-2 genomic sequences with near-perfect accuracy.
• A synthesized primer set demonstrated competitive performance, with 100% specificity and high sensitivity in patient samples.

Conclusions:

• The proposed AI-driven approach efficiently identifies unique viral sequences for diagnostic primer design.
• This method offers a rapid and accurate alternative for developing diagnostic tests, valuable for future pandemic response.
• The AI methodology can significantly reduce the time and data required for creating specific viral detection tools.