CNN-Based Detection of SARS-CoV-2 Variants Using Spike Protein Hydrophobicity
Mohammad Jamhuri, Mohammad Isa Irawan, Imam Mukhlash, Ni Nyoman Tri Puspaningsih · 2023
In the fight against the COVID-19 pandemic, it is crucial to quickly and accurately identify SARS-Co V-2 variants due to their ever-changing nature. In this study, we introduce a novel approach utilizing Convolutional Neural Networks (CNN) to classify the spike protein sequences of the virus, achieving an outstanding accuracy rate of 99.75%. For this method, we transformed a range of spike protein sequences, representing diverse SARS-CoV-2 variants, into images using the Kyte and Doolittle method to align with CNN input features. Comparative analyses with existing methodologies demonstrate the superior efficiency of our approach in terms of speed and precision. Such advancements in diagnostics play a fundamental role in shaping timely and informed public health strategies. Our research results showcase the potential of deep learning in tackling global health challenges and laying the groundwork for future innovations in virus diagnostics,