PUB - Publikationen an der Universität Bielefeld

Database searching and phylogenetic tree reconstruction are two major fields of computational sequence analysis. This thesis introduces a combination of both: a database search method that is based on phylogenetic trees - treesearch. A given protein family is described by its multiple alignment and its phylogenetic tree. A database sequence that is tested for membership in the family is tentatively inserted into that tree. The result of this operation determines how well the sequence fits into the family. The idea is realized in the distance based context of phylogeny. To assess the performance of the method in terms of sensitivity and selectivity, it is compared to profiles (ISREC pfsearch), two implementations of hidden Markov models (HMMER hmmsearch and SAM hmmscore), and to the family pairwise search (FPS) method. The comparison is based on a novel evaluation tool, which was also developed during this work. All methods are presented in a new unified functional framework of database searching. The analysis is complemented by extensive simulations. The treesearch idea is also transferred to the probabilistic context of phylogeny, which results in a generalization of established probabilistic methods such as pairwise sequence alignment, multiple sequence alignment, and profile searches.