Prediction Of Pandemic Risk For Coronavirus Based On Deep Learning

Since the end of 2019,COVID-19’s global pandemic has seriously damaged global economic and social activities.Through the two evolutionary mechanisms of point mutation and genome recombination,pandemic coronavirus is mutated from animal coronavirus,and continues to evolve and produce various variants in the process of transmission,which brings great challenges to epidemic prevention and control.In view of the current situation of COVID-19 and the future,the identification and prediction of the risk of animal origin coronavirus is a significant scientific issue.With the development of genome sequencing technology,a large number of coronavirus genome data have been collected in public data sets,which makes it possible to characterize and analyze the knowledge contained in them through deep learning methods.This paper intends to use the deep learning model to construct the prediction model related to virus infection phenotype,study the cross species infection of coronavirus and gene recombination between coronaviruses,and deeply understand the occurrence and development of virus.On this basis,a practical prediction tool is constructed to improve the prevention and control mechanism of infectious diseases when coronavirus infects people without crossing the species barrier or in the early stage of epidemic.The specific research contents are as follows:(1)coronavirus pandemic risk prediction model.The model takes coronavirus gene sequence as input,extracts local features using convolution neural network,extracts long-distance features using two-way gated loop network,and adds attention mechanism to improve the prediction weight of specific location,which is used to predict the impact of coronavirus gene mutation on the risk of cross species infection of animal derived coronavirus.At the same time,the effects of data preprocessing,pre training initialization and transfer learning on the performance of the model are studied.The experimental results show that the model can well capture the characteristics of coronavirus gene sequence and accurately predict the risk of human infection.(2)Coronavirus genome recombination prediction model.The model is adjusted based on the human risk prediction model of coronavirus cross species infection,and the gene sequences of two coronaviruses are input.The goal is to predict whether there is gene recombination relationship between different coronaviruses.At the same time,the effects of using location coding and multi head attention mechanism on improving the performance of the model are studied experimentally.The experimental results show that the model can well predict the gene recombination relationship between coronaviruses,and the optimized model can further improve the prediction effect.Coronavirus pandemic risk prediction model can effectively improve the early warning ability of early epidemic diseases.Further genome recombination prediction model can effectively monitor the evolutionary dynamics and potential pandemic risk of animal viruses in nature.This paper constructs the prediction software of the two models,which can be predicted directly through genomic data,which is convenient for relevant personnel in the field of medical health and animal health,and serves public health undertakings.