Mining negative sequential patterns to improve viral genomic feature representation and classification

Area of Science:

• Virology
• Bioinformatics
• Genomics

Background:

• Viruses are abundant and crucial in ecosystems, but also significant human pathogens.
• Current viral genome classification methods struggle with interpretability and accuracy on complex datasets.
• Existing models often rely on presence-based features, neglecting the importance of sequence absence.

Purpose of the Study:

• To introduce GeneNSPCla, a novel framework for viral classification using genomic negative sequential patterns (NSPs).
• To develop GONPM+, an improved algorithm for discovering biologically meaningful NSPs in genomic data.
• To enhance the accuracy and interpretability of viral genome identification.

Main Methods:

• GeneNSPCla utilizes Negative Sequential Patterns (NSPs) to extract absence-based features from RNA viral genomes.
• NSPs are converted into numerical vectors and integrated with supervised classifiers.
• The GONPM+ algorithm was developed for efficient and effective NSP discovery in genomic sequences.

Main Results:

• GeneNSPCla demonstrated improved accuracy in viral classification by incorporating absence signals.
• GONPM+ significantly outperformed previous negative and positive pattern mining algorithms.
• The average accuracy improvement with GONPM+ was 10.03% over the original NSP algorithm and 24.75% over positive pattern mining.

Conclusions:

• Incorporating absence-based sequential information offers a valuable new perspective for viral genome analysis.
• GeneNSPCla provides a more accurate and interpretable framework for classifying RNA viruses.
• The study highlights the potential of negative sequential patterns in bioinformatics and genomics.