Distance-Based Phylogenetic Method Research

Phylogenetic tree is a kind of typological structure for describing the sequence and relationship of species revolution. It is significant to construct phylogenetic tree in the biology field whose efficient and precise algorithms could yield to a great deal of practical value.We study the problem of constructing Phylogenetic tree in the paper. This problem is a NP complete problem whose algorithms are all heuristic. There are usually three main methods for the problem: (1) based-distance, (2) maximum parsimony, (3) maximum likelihood. We will merge maximum likelihood method to the based-distance algorithm process of constructing phylogenetic tree on the basis of advantages of both based-distance and maximum likelihood methods. There are two main phases in the process: firstly, we estimate phylogenetic distance among all the class groups in order to make sequence phylogenetic distance more precise, so we could construct more correct phylogenetic tree.Firstly, in the first phase of the algorithm, we study based-distance method and maximum likelihood method, change the traditional based-distance estimation method and enlarge the category of sequence alignment from double sequence to multiple sequence; we employ based-quartet maximum likelihood method to optimize phylogenetic models and use every possible topology with the maximum likelihood to estimate the distance between each two sequences so as to provide more correct distance matrix for constructing phylogenetic tree. Secondly, in the second phase, we mainly improve neighbor-joining for constructing phylogenetic tree, and there are two aspects in the process: One is distance matrix obtained on the first phase is not additve; and we introduce distance variance and covariance, renew the distance matrix by means of weighted computing formula and make the top structure built by the algorithm more correct. The other is to improve the greedy characteristics of neighbor-joining; Neighbor-joining so always aggregates two class units between which transformed distance is smallest as to lead to systematic bias. However, the experiment concludes that the couple whose transformed distance is smallest is not often the one between which the distance is not smallest in the real phylogenetic tree. Therefore, improved method always aggregates the two units whose transformed distance conforms to"neighbor", which does not always search the phylogenetic tree that has shortest sum of branch lengths, so the improved method heavily reduce the effect which systematic bias has on the real phylogenetic relationship. Then we testify our method's precision by experiments.Finally, we complete a phylogenetic tree construction system in terms of research results above.