Abstract:
Evolutionary tree (also called phylogenetic tree) is an interesting topic in theevolutionary study. This thesis proposes a new evolutionary tree construction algorithmthat constructs a tree with minimum total branch lengths for a given pairwise distance ofinput species. The problem of evolutionary tree, which is NP-complete, is transformedinto the Steiner's tree problem, i.e., to find the minimum length tree connecting all leafnodes. Ant Colony Optimization (ACO), a heuristic method, is applied to this problemwhich differs from the exact method of dynamic programming. To deal with thisproblem, a new procedure of tree construction, branch length calculation, and branchpoint selection are developed. ACO is used in the branch point selection. A new ACOparameter called distance-weighting parameter, K, is implemented to deal with distancecalculation between two branch points. The outputs of this algorithm are branch labelsand their length that are sufficient to construct an evolutionary tree.The developed algorithm can reduce the search space of the problem to polynomialcalculation time with complexity at the order of o(Nc.n4) ,where n is the number ofinput species and Nc is the number of iterations. When compared to the popularNeighbor Joining (NJ) method, the results from the new developed algorithm are better(shorter total distances) than NJ methods for 18 from 19 of the test cases.