A Novel Frequency Based Feature Extraction Technique for Classification of Corona Virus Genome and Discovery of COVID-19 Repeat Pattern

Abstract

Genome sequence regulates the life of all living organisms on earth. Genetic diseases cause genomic disorders and therefore early prediction of severe genetic diseases is quite possible by Genomesequence analysis. Genomic disorders refer to the mutation that is rearrangement of bases in the Genomeof an organism. Genome sequence analysis and mutation identification can help to classify the diseasedgenome which can be accomplished using Machine Learning techniques. Feature Extraction plays a crucialrole in classification as it is used to convert the Genome sequences into a set of quantitative values. In thisarticle, we propose a novel feature extraction technique called Frequency based Feature ExtractionTechnique which extracts 120 features from genome sequences for classification. In the current scenario,COVID-19 is the pandemic disease and Corona virus is the source of this disease. So, in this research work,we tested the proposed feature extraction technique with 1000 samples of Genome sequences of Coronavirus affected patients across the world. The extracted features were classified using both Machine Learningand Deep Learning techniques. From the results, it is evident that the proposed feature extraction techniqueperforms well with Convolutional Neural Network classifier giving an accuracy of 97.96%. The proposedtechnique also helps to find the most repeat patterns in the genome sequences. It is discovered that thepattern “TTGTT” is the most repeat pattern in COVID-19 genome