Reticulation in Evolution

Molecular phylogenetics, the study of reconstructing evolutionary trees, is a well-established field of scientific endeavor. However, in certain circumstances evolution is not completely tree-like. For example, a comparison of gene trees representing a set of present-day species and reconstructed for different genetic loci often reveals conflicting tree topologies. These discrepancies are not always due to missampling or difficulties in the gene tree reconstruction method, but rather due to reticulation events such as horizontal gene transfer (HGT) and hybridization. During an HGT event, a DNA segment is transferred from one organism to another which is not its offspring, whereas hybridization describes the origin of a new species through a mating between two different species. Both processes yield genomes that are mixtures of DNA regions derived from different ancestors. Consequently, evolutionary relationships between species whose past includes reticulation can often be better represented by using phylogenetic networks rather than trees. The main focus of this thesis is to develop new biologically motivated theoretical frameworks that provide insight into the extent to which reticulation events have influenced evolution. First, we have implemented the exact algorithm HybridNumber to compute the minimum number of hybridization events for two rooted binary phylogenetic trees. This approach is based on the notion of agreement forests and uses three rules that reduce the size of the problem instance, before calculating the hybridization number. We applied HybridNumber to a grass data set and analyzed the extent of hybridization. We also approached the question whether hybridization events have occurred relatively recently or in the distant past. Furthermore, since many biological data sets lead to reconstructed gene trees that are not fully resolved, we extended the above mentioned framework for rooted phylogenetic trees and showed that calculating the minimum number of hybridization events for two such trees is fixed-parameter tractable. Second, we present a new likelihood framework to estimate a rate of HGT for a set of taxa. To this end, we simulate an increasing number of HGT events on a species tree to obtain a tree distribution that can be used to estimate an HGT rate for a set of gene trees. This framework was applied to the COG (Clusters of Orthologous Groups of Proteins) data set and inaccuracies due to the gene tree reconstruction method were considered. Finally, we give a new result on how to speed up the exact calculation of the rooted subtree prune and regraft distance between two trees which is often used to model reticulation events and end with two interesting examples that give rise to questions for future research.