Transformada de wavelet discreta não decimada para o agrupamento de genomas de vírus das famílias coronaviridae e paramyxoviridae

Abstract

This work aimed to implement two forms of analysis of sequence similarities of two virus families, under the wavelet domain. Wavelets are commonly used when working with an extensive non-stationary database. The wavelet transform technique works with data in real time, and allows the time series to be decomposed into levels, thus allowing at each level of decomposition to increase the level of detail in the series, and thus observe details omissions, which cannot be observed in the original series. After decomposing the GC content of each of the sequences under study, two different forms of grouping were implemented in order to verify sequences with some level of similarity. Cluster analysis was carried out using penalized regression in the domain of lasso, ridge and elastic net penalties, and on the other hand, we also used the Hurst exponent implemented through 5 different techniques, namely: peng method, R analysis /S, aggregate variance, differentiated aggregate variance and by the method of absolute moments. At the end of the study, it can be concluded that weaker variants of the Coronaviridae family are associated with strains of the Paramyxoviridae family. And on the other hand, the elastic net (from the 1st to the 3rd level), the absolute moments method and the differentiated aggregate variance method performed better in relation to the other methodologies.