Classification of SARS-CoV-2 viral genome sequences using Neurochaos Learning.

Area of Science:

• Computational biology
• Machine learning
• Virology

Background:

• The rapid spread of SARS-CoV-2 necessitates advanced algorithms for rapid genome classification.
• Traditional machine learning models require extensive data, which is often unavailable during early disease outbreaks.
• Effective classification of viral genomes is critical for disease containment and public health.

Purpose of the Study:

• To introduce Neurochaos Learning (NL), a novel paradigm for classifying coronavirus genome sequences.
• To address the challenge of learning from limited training samples in viral outbreak scenarios.
• To evaluate the performance of NL against established machine learning classifiers.

Main Methods:

• Neurochaos Learning (NL) inspired by neuronal chaos and non-linearity.
• Multiclass classification of viral genomes (SARS-CoV-2, Coronaviridae, Metapneumovirus, Rhinovirus, Influenza).
• Leave-one-out cross-validation and comparison with K-nearest neighbours, logistic regression, random forest, SVM, and naïve Bayes.

Main Results:

• NL achieved average sensitivity, specificity, and accuracy of 0.998, 0.999, and 0.998, respectively.
• High average macro F1-score reported for classifying SARS-CoV-2 from SARS-CoV-1 with only one training sample per class.
• NL outperformed traditional machine learning methods in classification tasks with limited data.

Conclusions:

• Neurochaos Learning offers a promising solution for genome classification, especially with scarce training data.
• NL's ability to learn from minimal samples is vital for rapid response during disease outbreaks.
• Future applications include integrating NL with chaotic feature engineering for enhanced genome analysis.