GenoDense-Net: unraveling the genomic puzzle of the global pathogen.

Abstract

The respiratory system of humans is impacted by infectious and deadly illnesses like COVID-19. Early identification and diagnosis of this type of illness is essential to stop the infection from spreading further. In the present research, we presented a technique for determining the condition using COVID-19's current genome sequences employing the DenseNet-16 framework. We operated a network of already trained neurons before using a transfer learning method to prepare it according to our dataset. Additionally, we preprocessed the collected information using the NearKbest interpolation approach; then, we utilized Adam Optimizer to optimize our findings. Compared with special deep learning models like ResNet-50, VGG-19, AlexNet, and VGG-16, our approach produced an accuracy of 99.18%. The model was deployed on a platform with GPU support, which greatly decreased training time. Dataset size and the requirement for further validation are two of the study's limitations, despite the encouraging results. The current research showed how a deep learning approach may be useful to categorize the genome sequence of infectious disease like COVID-19 using the suggested GenoDense-Net architecture. The next step in this research project is conducting investigations in the clinic.