Robust and efficient COVID-19 detection techniques: A machine learning approach.

Area of Science:

• Virology
• Bioinformatics
• Computational Biology

Background:

• The Severe Acute Respiratory Syndrome-Coronavirus 2 (SARS-CoV-2) pandemic caused global disruption and significant mortality.
• While SARS-CoV-2 origins are debated, similarities with animal viruses are noted, yet its genome's microRNA interactions remain underexplored.
• MicroRNAs (miRNAs) play crucial roles in viral life cycles, but their specific involvement in SARS-CoV-2 requires further investigation.

Purpose of the Study:

• To develop a computational method for detecting SARS-CoV-2 precursor microRNAs (pre-miRNAs).
• To enhance the rapid identification of specific viral ribonucleic acid (RNA) sequences.
• To facilitate the design of targeted oligonucleotide-based therapies against SARS-CoV-2.

Main Methods:

• Utilized an artificial neural network (ANN) model for pre-miRNA detection.
• Employed random sampling on unbalanced datasets to mitigate class imbalance issues.
• Applied a matriculation ANN incorporating accuracy curves, loss curves, and confusion matrices for performance evaluation.

Main Results:

• The proposed ANN model demonstrated a high estimated accuracy of 98.24% in detecting SARS-CoV-2 pre-miRNAs.
• The method effectively identified target regions within the viral RNA.
• Performance comparison showed the ANN model outperforming classical machine learning approaches.

Conclusions:

• The developed ANN-based method offers a highly accurate and rapid approach for SARS-CoV-2 pre-miRNA detection.
• This technique can significantly improve the recognition of the SARS-CoV-2 genome sequence.
• The findings support the design of novel oligonucleotide-based drugs targeting the virus's genetic structure.