Alignment-Free Viral Sequence Classification at Scale.

Area of Science:

• Computational Biology
• Genomics
• Bioinformatics

Background:

• Next-generation sequencing (NGS) generates vast amounts of nucleotide data, overwhelming traditional alignment-based sequence comparison tools like BLAST.
• The computational demands of alignment-based methods limit their scalability for classifying massive viral datasets.
• There is a critical need for efficient and accurate computational tools for viral sequence classification in the era of big data.

Purpose of the Study:

• To evaluate alignment-free (AF) methods as scalable and rapid alternatives for viral sequence classification.
• To identify AF techniques that maintain high accuracy and efficiency on extremely large viral sequence datasets.
• To assess the suitability of AF methods for classifying diverse viral genomes, including SARS-CoV-2, dengue, and HIV.

Main Methods:

• Six established alignment-free techniques were utilized to extract feature vectors from viral genomes.
• Random Forest classifiers were trained using these feature vectors for sequence classification.
• Models were trained and tested on a large dataset of 297,186 SARS-CoV-2 sequences across 3502 lineages, and validated on dengue and HIV datasets.

Main Results:

• Alignment-free classifiers achieved high accuracy: 97.8% for SARS-CoV-2, 99.8% for dengue, and 89.1% for HIV.
• Word-based alignment-free methods effectively represent viral sequences, even with high-dimensional class structures.
• AF techniques demonstrated significantly faster processing times compared to alignment-based methods.

Conclusions:

• Alignment-free methods provide a scalable and efficient solution for viral sequence classification.
• These techniques are suitable for handling large-scale genomic datasets using modest computational resources.
• AF methods offer practical advantages for rapid viral classification, addressing the limitations of traditional approaches.