Unsupervised machine learning framework for discriminating major variants of concern during COVID-19.

Area of Science:

• Genomics
• Virology
• Computational Biology

Background:

• The rapid evolution of SARS-CoV-2, driven by high mutation rates, led to variants like Delta and Omicron with increased transmissibility and severity.
• These variants significantly impacted global health systems, economies, and travel.
• Unsupervised machine learning offers powerful tools for analyzing complex, unlabeled biological data.

Purpose of the Study:

• To develop and present a framework using unsupervised machine learning to differentiate and visualize relationships between major COVID-19 variants based on their genomic sequences.
• To assess the framework's capability in identifying mutational differences among variants and geographically.

Main Methods:

• The framework employs k-mer analysis of RNA sequences.
• Dimensionality reduction techniques including Principal Component Analysis (PCA), t-distributed Stochastic Neighbor Embedding (t-SNE), and Uniform Manifold Approximation Projection (UMAP) are used for visualization.
• Agglomerative hierarchical clustering and dendrograms are utilized to illustrate mutational variations.

Main Results:

• The proposed framework successfully distinguishes between major COVID-19 variants using genomic sequence data.
• Mutational differences among variants of concern and country-specific variations for Delta and Omicron were visualized.
• The framework demonstrated effectiveness in differentiating variants and has potential for identifying new ones.

Conclusions:

• The unsupervised machine learning framework provides an effective method for discriminating between known COVID-19 variants.
• The approach shows promise for the early detection and characterization of emerging SARS-CoV-2 variants.
• This computational tool can aid in monitoring viral evolution and informing public health strategies.