Alignment-free viral sequence classification at scale.

Area of Science:

• Bioinformatics
• Computational Biology
• Genomics

Background:

• Next-generation sequencing (NGS) generates vast amounts of nucleotide data, necessitating efficient sequence comparison tools.
• Traditional alignment-based methods (e.g., BLAST) struggle with scalability for massive datasets.
• Alignment-free (AF) methods present a promising, computationally efficient alternative for sequence analysis.

Purpose of the Study:

• To evaluate the efficacy of AF methods for large-scale viral sequence classification.
• To identify AF techniques that balance accuracy and computational efficiency.
• To assess the performance of AF methods across diverse viral datasets.

Main Methods:

• Six established AF techniques were utilized to generate feature vectors from viral genomes.
• Random Forest classifiers were trained using these feature vectors.
• Models were validated on extensive SARS-CoV-2, dengue, and HIV sequence datasets.

Main Results:

• AF classifiers achieved high accuracy: 97.8% for SARS-CoV-2, 99.8% for dengue, and 89.1% for HIV.
• Word-based AF methods effectively represent viral sequences, even with high dimensionality.
• Demonstrated robustness and scalability across different viral genomic datasets.

Conclusions:

• AF methods provide a practical and efficient solution for viral sequence classification.
• These techniques offer significant speed advantages over alignment-based methods.
• AF approaches enable sequence classification using limited computational resources.