Reconstructing phylogenies based on distance between transforms

Details

Supervisors

Catanzaro, Daniele

Faculty

Louvain School of Management

Degree label

Master [60] en sciences de gestion

Abstract

For the last decades, phylogenetics have become more and more important in genetics researches and of prime interest for DNA sequences analysis. Alignment-free methods have overtaken multiple sequence align-ment methods, quite complex and memory consuming. But several of these alignment-free methods are still complex and do not use all the information available (loss of information content). To overcome this issue, new methods have been developed using content of the signals in another domain, using transforms preserving the energy levels, and so, the information). Transforms such as Fourier or Ramanujan are now used for that purpose. DNA sequences (composed of the four types of nucleotides) are often represented using binary code to be usable. The goal of this paper is to introduce new transforms and distances (applied on transforms), not used in this context before, and more convenient for binary signals or to compute distances between spectra. The accuracy of the new methods is assessed using phylogenetic trees to evaluate the quality of the new distance computation. They are built based on artificial and real datasets. The results show that some combinations of new transforms and metrics produce accurate distance computation and phylogenetic trees.