Inferring phylogenetical tree by using hierarchical self organizing maps

Abstract

In biology, inferring phylogenetical tree is an attempt to describe the evolutionary history of today’s species with the aim of finding their common ancestors. Specifically in molecular biology, it is used in understanding the evolution relationships between proteins or DNA sequences. Inferring phylogenetical tree can be a very complicated task since even for the input data having thirty sequences, the best tree must be chosen among 1036 possible trees. In order to find the best one in a reasonable time, various hierarchical clustering techniques exist in the literature. On the other side, it is known that Self Organizing Maps (SOM) are very successful in mapping higher dimensional inputs to two dimensional output spaces (maps) without having any priori information about input patterns. In this study, SOM are used iteratively for tree inference. Two different algorithms are proposed. First one is hierarchical top-down SOM method which constructs the tree from the root to the leaves. Second one uses a bottom-up approach that infers the tree from the leaves to the root. The efficiency of Hierarchical SOM is tested in terms of tree topology. Hierarchical SOM gives better results than the most popular phylogeny methods, UPGMA and Neighbor-joining. Also this study covers possible solutions for branch length estimation problem.