Sequence alignment and phylogenetic inference

by Roland Fleissner

This thesis deals with some of the problems arising at the junction between sequence alignment and phylogeny After giving a brief overview over some basic concepts and methods of phylogenetic analysis in chapter 1, we use the concept of parametric sequence alignment to show that substitution rates cannot be estimated reliably from optimal sequence alignments (chapter 2). Therefore, we outline a simple least-squares method which leads to better distance estimates in a large area of the parameter space (chapter 2). The simulations described in chapter 3 exemplify the problems that may arise when one reconstructs trees from multiple alignments. We observe a drastic reduction of the efficiency of neighbor-joining even if the sequences are reasonably long. We also see that multiple alignment programs may force their implicit assumptions upon the sequences under study, thus being the cause of erroneous trees and erroneous support values. Chapter 4 gives an introduction into the application of insertion and deletion models. These models are then used to simultaneously reconstruct phylogenies and multiple alignments (chapter 5). Finally, we describe in the appendices the usage and some implementation details of the main programs that were written for this thesis as well as some possible generalizations of the models applied.

137 pages, Paperback

First published May 30, 2004