The evolution and development of eyespot and non-eyespot colour pattern elements was studied in Vanessa butterflies using a phylogenetic approach. A Bayesian phylogeny of the genus Vanessa was reconstructed from 7750 DNA base pairs from 10 genes. Twenty-four non-eyespot and forty-four eyespot color pattern elements from the Nymphalid ground plan were defined and studied and their evolutionary history was traced on the Vanessa phylogeny. Ancestral character states were predicted and the direction of evolutionary changes was inferred for all characters. Five serially arranged eyespots were predicted for the ancestral Vanessa on all wing surfaces. Homologous eyespot and non-eyespot characters on the surfaces of the forewing were more similar than those on the surfaces of the hindwing. Homologous eyespot characters on the dorsal surfaces of fore and hindwings show more similarities than the ventral surfaces, in contrast to what was found for non-eyespot characters. Independent Contrast analysis was also used to study correlations between eyespot characters. Independent Contrast analysis revealed significant correlations between eyespots 2 and 5 and eyespots 3 and 4 on all wing surfaces. This consistency among highly variable eyespot characters suggested a structural hypothesis: the existence of a Far-Posterior (F-P) compartment boundary and organizer could be responsible for the observed correlations. This hypothesis was tested in several ways. First, examination of wing patterns across species from all families of butterflies revealed correspondence between wing cells 1 and 4 and between cells 2 and 3. Second, evaluation of spontaneous mitotic clones in butterflies and moths reveals a peak abundance of clonal boundaries along the vein dividing wing cells 2 and 3. Finally, experimentally generated FLP/FRT mitotic wing clones produced in Drosophila, reveal a clonal boundary posterior to the L5 wing vein, which is homologous to the vein dividing wing cells 3 and 4 in butterflies. Collectively, this suggests the existence of an additional compartment boundary associated with an organizer in wing cell 3 responsible for patterning the posterior portion of insect wings. A model is proposed that predicts that the wing developmental compartment boundaries produce unique combinations of gene expression for each wing sector, permitting eyespot individuation. / February 2016

In den letzten Jahren haben Anzahl und Umfang verfuegbarer Datensaetze stark zugenommen, wodurch die Entwicklung von Methoden zur Entdeckung von Wissens in den Daten zu einer grossen Herausforderung geworden ist. Waehrend dabei sonst Effizienzfragen im Vordergrund standen, wurde in juengerer Zeit auch die temporale Dimension der Daten einbezogen. Es wurden Methoden erarbeitet, die der Pflege des entdeckten Wissens dienen. Diesen Techniken liegt die Idee zugrunde, dass Daten oft ueber einen langen Zeitraum gesammelt werden. Damit sind sie den gleichen Aenderungen ausgesetzt wie die Realitaet. Aendern sich aber die Daten, ist auch mit Aenderungen in den Analyse-Ergebnissen zu rechnen. Es genuegt aber nicht, nur die Aktualitaet der Ergebnisse sicherzustellen. Vielmehr ist es notwendig, auch ihre Entwicklung im Zeitverlauf zu erfassen. In dieser Arbeit wird Wissensentdeckung als kontinuierlicher Prozess verstanden. Daten werden ueber einen potentiell langen Zeitraum gesammelt und in bestimmten Zeitabstaenden analysiert. Jede Analyse liefert eine Menge von Mustern, die in einer Regelbasis erfasst und deren Entwicklung aufgezeichnet wird. Ausgangspunkt ist ein temporales Datenmodell, das den Inhalt von Mustern und ihre statistischen Eigenschaften abbildet. Darauf aufbauend, wird ein umfassendes Bezugssystem fuer die Ueberwachung und Analyse der Entwicklung entdeckten Wissens entwickelt, das die vielen verschiedenen Facetten der Evolution von Mustern integriert und die Erkennung von Trends erlaubt. Dieses Bezugssystem ermoeglicht es, verschiedene Arten von Musteraenderungen nach qualitativen, quantitativen und temporalen Kriterien erkennen und bewerten zu koennen, andererseits gestattet es, die temporalen Eigenschaften der gefundenen Zusammenhaenge als Kriterium fuer ihre Relevanz zu nutzen und die Ursachen der beobachteten Aenderungen zu bestimmen. Im Rahmen zweier Fallstudien wurden die vorgestellten Konzepte einer eingehenden Ueberpruefung unterzogen. / Over the past years the number and size of datasets have grown significantly. This has stimulated research into the development of techniques for the discovery of knowledge in this data. Traditionally the emphasis has been on criteria such as performance and scalability; in recent years, however, the temporal dimension of the data has become a focus of interest. Methods have been developed that deal with the maintenance of the discovered knowledge. These approaches are based on the assumption that the data is collected over a long period of time and, thus, affected by the same changes as the aspects of reality captured in the data. Hence, changes to the data will also be reflected in changes to the results of analysing the data. Therefore, it is not sufficient to consider only the non-temporal aspects of the knowledge, rather it becomes a necessity to also consider the development of identified patterns over time. In this work, knowledge discovery is considered to be a continuous process: data is collected over a period of time and analysed at specific time intervals. Each analysis produces a set of patterns which are stored in a rule base and monitored based on their statistical properties. Using a temporal data model which consists of both the content of a pattern and its statistical measurements, a general framework for monitoring and analysing the development of the discovered knowledge is proposed. Integrating the many different facets of pattern evolution, the model also provides for trend recognition. The framework is used to detect and assess different types of pattern change with respect to their qualitative, quantitative and temporal aspects. In addition, it permits the usage of the temporal properties of patterns as criterion for their relevance and enables the application expert to determine the causes of pattern change. Two case studies are presented and discussed which examine the eligibility of the proposed concepts thoroughly.