Efficient Approximate Kernel Based Spike Sequence Classification.

Area of Science:

• Bioinformatics
• Computational Biology
• Machine Learning

Background:

• Machine learning models require sequence similarity metrics for classification and clustering.
• Exact k-mer matching methods are computationally expensive, limiting scalability.
• Approximate methods offer scalability but lack domain specificity.

Purpose of the Study:

• To improve approximate kernel performance for coronavirus sequence analysis.
• To enhance predictive accuracy for classifying COVID-19 variants.
• To integrate domain knowledge and efficient preprocessing into sequence similarity computation.

Main Methods:

• Utilized minimizers for efficient preprocessing of sequences.
• Incorporated information gain to compute domain knowledge.
• Developed an improved approximate kernel for coronavirus spike protein sequences.
• Applied various classification and clustering algorithms for evaluation.

Main Results:

• The proposed method significantly improved kernel performance compared to baseline and state-of-the-art approaches.
• Enhanced predictive accuracy in classifying coronavirus variants (e.g., Alpha, Beta, Gamma).
• Demonstrated improved performance across multiple evaluation metrics on two datasets.

Conclusions:

• The enhanced approximate kernel offers a more accurate and scalable solution for coronavirus sequence classification.
• Integrating domain knowledge and efficient preprocessing is crucial for specialized sequence analysis.
• The approach shows promise for applications in infectious disease research and healthcare.