Hydrophobicity signal analysis for robust SARS-CoV-2 classification

Abstract

Rapid and accurate classification of viral pathogens is critical for effective public health interventions. This study introduces a novel approach using convolutional neural networks (CNN) to classify SARS-CoV-2 and non-SARS-CoV-2 viruses via hydrophobicity signal derived from DNA sequences. Conventional machine learning methods grapple with the variability of viral genetic material, requiring fixed-length sequences and extensive preprocessing. The proposed method transforms genetic sequences into image-based representations, enabling CNNs to handle complexity and variability without these constraints. The dataset includes 8,143 DNA sequences from seven coronaviruses, translated into amino acid sequences and evaluated for hydrophobicity. Experimental results demonstrate that the CNN model achieves superior performance, with an accuracy of over 99.84% in the classification task. The model also performs well with extended sequence lengths, showcasing robustness and adaptability. Compared to previous studies, this method offers higher accuracy and computational efficiency, providing a reliable solution for rapid virus detection with potential applications in bioinformatics and clinical settings.