A deep learning-based model for mutation rate prediction of COVID -19 using genomic sequences
Rajit Nair, Maki Mahdi Abdulhasan, Hameed Hassan Khalaf, Ashraf Mohammed Shareef
Abstract
Purpose: SARS-CoV-2 has manifested itself as a contagion to have caused a global pandemic with a new coronavirus known primarily as COVID-19. This viral RNA virus is now immobilizing the world. More than 148,480,035 were already infected as of April 25th, 2021, and 3,133,637 died. A fact has been ascertained that SARS-CoV-2 mutates at a rapid pace leading to its enhanced virulence which not only renders it to be highly transmissible but also fatal than earlier variants. A precise determination of mutation rate is needed to understand pathogenecity of virus and to analyze an emerging disease’s risk.Method: This research examines the rate of mutation in the entire genomic sequence deciphered from both oro-pharyngeal and naso-pharyngeal samples from sources across the globe. The data set is analyzed for determining the mutation of codon and nucleotides individually. Moreover, the calculated rate of mutation is classified for four diverse countries based on the data set size: Europe, North America, India, and Oceania. A significant quantity of adenine (A) and thymine (T) bases have been found to mutate into further nucleotides in the entire sequence, however, codons do not often mutate like nucleotides. An approach based on a deep neural network is suggested to analyse the risk during virus mutation.Results: The proposed model has obtained Root Mean Square Error (RMSE) of 0.05 on the training dataset and 0.03 on the testing dataset and this value is much better than the existing models. The nucleotide mutation rate of 500th patients has been expected in the future through this training and testing process.Conclusion: An increase of approximately 0.1 percent can be observed, for mutating nucleotides resulting from transversion of nucleobases.