Análises filogenéticas são bastante utilizadas para a compreensão das relações existentes entre objetos biológicos, beneficiando as investigações em vários campos das ciências da vida. Vários métodos computacionais para reconstruir filogenias tem sido desenvolvidos. Em geral, os métodos que fornecem filogenias mais confiáveis, requerem significativamente maior tempo computacional, restringindo a aplicação deles a conjuntos de dados relativamente pequenos. Por outro lado, a utilização de conjuntos de dados maiores é fundamental para proporcionar uma amostragem que seja suficiente, para restringir as incongruências na identificação de clados em uma filogenia. Este trabalho propõe uma abordagen (denominada CladeNet) de reamostragem de filogenias, obtidas por algoritmos relativamente eficientes, a fim de melhorar a identificação de clados. Experimentos com sete conjuntos de dados, que variam de dezenas a centenas de sequências de DNA mostram que, em geral, clados encontrados pela abordagem proposta tornam-se mais confiavéis, conforme os tamanhos dos conjuntos de sequências aumentam, com um moderado aumento do tempo computacional relativamente moderado. Além disso, o CladeNet é um método que também inova ao identificar clados de forma automáticamente por meio de um algoritmo de identificação de comunidades em redes. / Phylogenies are useful for understanding relationships among biological objects, benefiting investigations in various fields of life sciences. Several computational methods for reconstructing phylogenies have been developed. In general, methods that provide more reliable phylogenies require significantly larger computing time, constraining their application to relatively small datasets of objects. On the other hand, the use of larger datasets is fundamental to provide enough samples in order to reduce incongruence in clade identification from a phylogeny. This work proposes an approach of resampling phylogenies (called CladeNet) obtained from relatively efficient algorithms, in order to improve clade identification. Experiments with seven datasets with sizes varying from dozens to hundreds of DNA sequences show that, in general, clades found by the proposed approach are more reliable as the dataset sizes augment, with relatively moderate increase of computing time. Moreover, CladeNet is a new method for identifying clades in an automatic way by means of community detection algorithm for networks.

Análises filogenéticas são bastante utilizadas para a compreensão das relações existentes entre objetos biológicos, beneficiando as investigações em vários campos das ciências da vida. Vários métodos computacionais para reconstruir filogenias tem sido desenvolvidos. Em geral, os métodos que fornecem filogenias mais confiáveis, requerem significativamente maior tempo computacional, restringindo a aplicação deles a conjuntos de dados relativamente pequenos. Por outro lado, a utilização de conjuntos de dados maiores é fundamental para proporcionar uma amostragem que seja suficiente, para restringir as incongruências na identificação de clados em uma filogenia. Este trabalho propõe uma abordagen (denominada CladeNet) de reamostragem de filogenias, obtidas por algoritmos relativamente eficientes, a fim de melhorar a identificação de clados. Experimentos com sete conjuntos de dados, que variam de dezenas a centenas de sequências de DNA mostram que, em geral, clados encontrados pela abordagem proposta tornam-se mais confiavéis, conforme os tamanhos dos conjuntos de sequências aumentam, com um moderado aumento do tempo computacional relativamente moderado. Além disso, o CladeNet é um método que também inova ao identificar clados de forma automáticamente por meio de um algoritmo de identificação de comunidades em redes. / Phylogenies are useful for understanding relationships among biological objects, benefiting investigations in various fields of life sciences. Several computational methods for reconstructing phylogenies have been developed. In general, methods that provide more reliable phylogenies require significantly larger computing time, constraining their application to relatively small datasets of objects. On the other hand, the use of larger datasets is fundamental to provide enough samples in order to reduce incongruence in clade identification from a phylogeny. This work proposes an approach of resampling phylogenies (called CladeNet) obtained from relatively efficient algorithms, in order to improve clade identification. Experiments with seven datasets with sizes varying from dozens to hundreds of DNA sequences show that, in general, clades found by the proposed approach are more reliable as the dataset sizes augment, with relatively moderate increase of computing time. Moreover, CladeNet is a new method for identifying clades in an automatic way by means of community detection algorithm for networks.

This thesis focuses on phylogenetic relationships among some of the major lineages in Silene subgenus Behenantha (Caryophyllaceae) using DNA sequences from multiple, potentially unlinked gene regions from a large taxonomic and geographic sample. Both traditional phylogenetic analyses and a strategy to infer species trees and gene trees in a joint approach are used. A new strategy to optimize species classifications, based on the likelihoods of the observed gene trees, is presented. Silene latifolia, S. dioica and the other dioecious species previously classified in section Elisanthe are not closely related to the type of the section (S. noctiflora). The correct name for the group of dioecious species is section Melandrium. The chloroplast DNA data presented indicate a geographic, rather than a taxonomic, structure in section Melandrium. The nuclear genes investigated correlate more to the current taxonomy, although hybridization has likely been influencing the relationships within section Melandrium. Incongruence between different parts of the gene SlXY1 in two Silene lineages is investigated, using phylogenetic methods and a novel probabilistic, multiple primer-pair PCR approach. The incongruence is best explained by ancient hybridization and recombination events. A survey of mitochondrial substitution rate variation in Sileneae is presented. Silene section Conoimorpha, S. noctiflora and the closely related S. turkestanica have elevated synonymous substitution rates in the mitochondrial genes investigated. Morphological and phylogenetic data reject that the Californian S. multinervia should be treated as a synonym to the Asian S. coniflora, as has previously been suggested. Furthermore, none of the genes investigated, or a chromosome count, support the inclusion of S. multinervia in section Conoimorpha. Data from multiple genes suggest that S. noctiflora and S. turkestanica form a sister group to section Conoimorpha. The calyx nervature, which is a potential synapomorphy for S. multinervia and section Conoimorpha, may be explained either by parallelism or by sorting effects.

The effect of a free-living stage in host-parasite coevolution: a skink mite phylogenetic study in New Zealand. During the last decade, phylogenetic trees have even been used to compare ecologically related taxa such as parasites and their hosts, and are used to determine their level of coevolution or reciprocal adaptation in time. Diverse coevolutionary events have been detected for this ecological association, where generally the parasite has been regarded as one that feeds exclusively on the host and is likely to cospeciate with it. A different coevolutionary pattern might occur when the parasite has a free-living stage in its life cycle, in which the parasite may have the opportunity to abandon its host and successfully colonise a new species (host-switching) making cospeciation less likely. Many New Zealand skinks are infested with a parasitic mite, Odontacarus sp. (Prostigmata: Leeuwenhoekiidae), which becomes free-living as an adult. The genetic variation of these mites found on four hosts was analyzed for host- parasite coevolutionary events. The hosts were the McCann’s skink and the common skink in coastal Birdling Flat, Canterbury, plus these species and the Grand and Otago skinks in Macraes Flat, Central Otago, South Island, New Zealand. The genetic variation of fast evolving nuclear Internal Transcribed Spacers 2 and mitochondrial Cytochrome c Oxidase I in Odontacarus mites found on these hosts was determined by PCR and DNA sequencing and phylogenetic trees were built using the computer programs PAUP*4 and MrBayes 3. The results show that mite haplotypes only had a significant geographical division and no host-related differences. In Birdling Flat, the COI haplotypes were represented in two groups that infested both regional hosts and had 5.7 % divergence. The same individual mites belonged to a single ITS 2 haplotype, thus indicating a historical geographical division between two populations that now interbreed successfully. The Macraes Flat mites were divided into two COI haplotypes with 2.4% divergence and internal nodes, which showed greater genetic variability than the Birdling Flat populations. The Macraes Flat mites formed two ITS 2 haplotypes with 6% divergence. This greater geographical structure of the Otago mites is probably due to the older age of the mainland area compared to the recently exposed coastal locality of Birdling Flat. The COI haplotypes from the two different regions had a mean distance of 15.5%, with an earlier divergence time than that known for the hosts. For both genes, the haplotypes from different regions had 100% bootstrap support and the parasite showed no host specificity. Mites of the different COI and ITS haplotypes were found on most of the host species that were sampled in Canterbury and Otago. The results of this study suggest that a free-living stage in a parasite’s life cycle can favour coevolutionary events such as inertia (failure to speciate) and host-switching, probably as a result of resource-tracking of the parasite. NB: Electronic files contained on CD to accompany print copy are not included with this version of the thesis.