An automated ultra-fast, memory-efficient, and accurate method for viral genome classification

Area of Science:

• Virology
• Bioinformatics
• Computational Biology

Background:

• Virus classification is crucial for virology and epidemiology.
• Existing methods face performance and computational challenges, especially for Influenza and HIV.
• Need for efficient and accurate automated virus classification tools.

Purpose of the Study:

• To design an automated classifier, PC-mer, to overcome existing limitations in virus classification.
• To reduce feature dimensionality and improve training speed.
• To enhance classification performance across various virus families.

Main Methods:

• Developed PC-mer, an automated classifier based on k-mer and physicochemical characteristics of nucleotides.
• Reduced feature count significantly compared to traditional k-mer methods (by 2^k times).
• Maintained a constant feature count irrespective of sequence length, unlike integer and one-hot encoding.

Main Results:

• Achieved an average training speed increase of 17.93 times.
• Demonstrated improved classification performance for multiple virus families.
• Significantly reduced processing complexity and memory bottlenecks.

Conclusions:

• PC-mer offers a computationally efficient and accurate approach to virus classification.
• The method effectively addresses performance bottlenecks in classifying challenging virus types.
• PC-mer represents a significant advancement in automated virus identification and analysis.