Modular hierarchical modeling formalisms distinghuish clearly between simulation algorithms and model definitions and offer well-defined means to construct models flexibly. At the same time they are particularily aimed at a centralized development of models. In contrast, simulation-based integration approaches clearly separate interface descriptions from model implementations, but offer less support for constructing models and experimenting with them. A carefull combination of concepts from the area of component-oriented software development, model-driven development and service-oriented architectures allows to combine the advantages of model-based and simulation-based composition approaches. This thesis conceives, defines, and realizes a platform for the composition of simulation models. The platform allows to associate model implementations, defined in specific formalisms, with formalism-independent interface definitions and to deploy models as parametrizable components. Interface definitions may be stored as XML documents in databases to search, select, and analyze composition candidates. A meta model formalizes the description means of the component platform, such that the refinement of interfaces into model implementations and the compatibility of interfaces can be checked automatically. The composition tool CoMo (Component Models) realizes the platform definition in the Java programming language. CoMo allows to derive platform-specific simulation models from platform independent compositions. The integration of simulation platforms is demonstrated on the example of the existing simulation system James II. The developed composition platform is used in a simulation study to evaluate service trading mechanisms in mobile ad-hoc networks.

Applying Information Retrieval (IR) techniques on model retrieval is gaining more importance with the fast growing number of computational biological models (hereafter named bio-models). Several work groups develop and use different formalisms and interchange formats for bio-models and store those models in various model data bases and repositories.

Mostly, models are lost to the community because they lack a proper annotation or documentation, use an uncommon or new formalism. Overcoming these losses would enable models to be shared and reused between communities. To render sharing and reuse of bio-models possible the ability to find bio-models (i.e. to efficiently retrieve them from a repository) is mandatory. However, the current state-of-the-art is to provide the user with an unranked list of models for his query; as currently done in BioModels Database or in the CellML Model Repository.

Searching for biological models is hardly sufficient by querying raw model code of a certain interchange format. Presumably, it is worthwhile to include additional information about a model:

1.  What is known about a model's constituents?
2.  What kind of relations exist between the constituents?
3.  How does a model behave under certain conditions?

Most of these questions can be answered by annotations. Thus, a thorough annotation of bio-models enhances the semantic description of the modeled system by far and can give a contribute to sophisticated IR techniques. Efforts for enhanced model annotation, such as the "Minimum Information Requested in the Annotation of Biochemical Models" (MIRIAM) approach, are already applied to some modeling formalisms, e.g. the "Systems Biology Markup Language" (SBML). The main question for this phd work is how the information available about a bio-model can be gathered, processed and analyzed, and stored in a manner to efficiently retrieve bio-models. Furthermore, it is a research question how to rank the retrieved models with respect to a user's query and needs.

Within this project, the formalism ML-DEVS for explicitly describing multi-level models has been already developed. A modeling study on RNA structure folding revealed that it is possible to describe cell-biological and molecular systems with ML-DEVS. However, it also turned out that the language concept which is based on the metaphor of reactive systems with asynchronous communication patterns hampers modeling of biochemical reactions. Therefore, the goal of this project is to incorporate the previously achieved results for multi-level modeling - including those results gained by Mathias John - into a formal language with well-defined syntax and semantics. Thereby, the syntax shall be defnined in an intuitive way, so that the language can be easily learned by domain experts from biology. A rule-based approach following the paradigm of chemical reaction definitions seems to be promising. Furthermore, the syntax shall explicitly support the description of hierarchical model structures and interactions between different levels. The semantics shall follow the well-established concept of continuous time Markov chains. To examine the applicability of the formal language concept for cellbiological systems, the project shall further include a detailed modeling study.

Intelligent teaching and training systems can extend education in Biology, Chemistry and Medicine. They are a meaningful supplement to traditional lectures and courses, as they allow interactive training with realistic training cases from an early stage of education. Computer-based experimentation, based on techniques of modeling and simulation, are important part of state of the art research projects in Biology and Chemistry, and also in Medicine. However, this is in most cases not reflected in the current curricula. A case-based teaching and training system for teaching modeling and simulation, i.e. computer based experimentation, e.g. to Biologists, will be developed in this project. The teaching and trainings system, realized as an Intelligent Tutoring System (ITS), is based upon bio-chemical expert knowledge in combination with knowledge about (quantitative) modeling. The ITS functionality is extended by a high potential simulation system (JAMES II). Using this system, students and researchers in the interdisciplinary field of systembiology should be able to train different phases of computer-based experimental design and experiment execution, starting from formalism selection, to model development, model verification, parameterization of models, simulation selection and finally simulation execution and evaluation -- all of this based on close-to-real-life scenarios.

The goal of this project will be to develop modeling concepts for describing structural and spatial change in regenerative systems. Thereby, modeling formalisms that deal with structural and spatial changes shall be analyzed and more recent developments explored. The developed modeling concepts should support an effective description of the interesting phenomena, support an visualization and allow an efficient execution. The evaluation will be based on concrete cell-biological applications where space and structure plays an important role.

There are different storage models for XML since they highly depend on the application field and the usage of XML. Other influences are aspects like distribution of document collections, re-use of documents, query and update classes, evolution and adoption of structures (schemas) used. The challenges are the design and deployment of a distributed XML repository and efficiently conceived query and update processing facilities. The retrieval of model components for re-use is one main issue which has to get worked out together with the modelling and simulation group. The retrieval should be based on content or by example, i.e. query by example or query by structure. Therefore, techniques should be employed which are known from the field of structure/schema extraction, structure mining. Additionally, schema matching algorithms are needed as well as query processing techniques for semi-structured data and content based retrieval methods. The combination and selection of the specialized storage and query processing methods rely on knowledge from the application domain. This knowledge can be extracted and derived from existing standards (e.g. SBML markup language), well-documented database applications from system-biology and their schemas. New evolving technologies like ontologies from OBO (Open biomedical ontologies) and other tools from the semantic web should be regarded and employed as well.

When modeling, simulating, and evaluating regenerative systems, huge volumes of data have to be considered. In these scenarios it makes sense to take advantage of visualization methods. Such methods provide not only intuitive overviews, but can also reveal relations within the data.

Moreover, visualization is useful to prove or disprove hypotheses about the data. The goal of this project is to develop intuitive visual methods to support:

• the generation and analysis of hypotheses about derived data
• the generation of methods and models, as well as
• the evaluation of hypotheses, methods, and models.

On the one hand, this requires contributions to the young field of Bio-Informatics. Previous work on visualizing gene expressions was conducted in cooperation with the team of Prof. Rolfs. However, to cope with huge heterogeneous data, new solutions have to be developed.

On the other hand, visualization techniques to support the generation of hypotheses and models are rare in literature. New concepts have to be devised to support deeper understanding of the foundations of regenerative systems.

Visual data mining describes the general procedure to explore larger amounts of data. The idea behind visual data mining is to combine visual and analytical methods to extract salient features of the data set at hand. Today’s tools incorporate mining methods that work on data values (clustering, association analysis, etc.). The analysis of structural relations has not yet been in the focus of research. On the other hand, a variety of techniques for the visualization of structures are known in literature. The goal of this project is to investigate how analytical algorithms (e.g., as known from graph theory) can be used to drive (in the sense of pre-calculations and abstractions) the visualization of larger model structures. This is particularly useful to visualize interactions of proteins and to support the exploration of biological models.

The dissertation project examines the design of computer games with learning content from three different viewpoints: computer science, teaching psychology and media science. Computer-based learning games are expected to boost a learner's intrinsic motivation, to enable playful trial and error, and to facilitate knowledge transfer. Their success, however, largely depends on the meaningful integration of learning content into the mechanisms of the game. The same holds true for the next generation of adaptive (towards the learner's needs) learning games. One of the assumptions of this work is that game mechanisms taken from computer role playing games are particularly useful to transfer learning content. Based on this premise a reference software architecture will be instantiated with a special learning paradigm for science education: virtual laboratories. Based on design patterns these interactive experiments are embedded into game quests. In order to develop and validate the architecture, learning paradigm, and the design patterns, an accompanying learning game in the domain of marine sciences is to be developed.

In this PhD, visual methods have been developed in the context of climate impact research to support the construction and evaluation of models. Furthermore, principal investigation has been done in the field of visualization design in order to automatically generate adequate, problem-oriented representations. It is to be expected that the developed solutions can also be adapted and beneficially applied to the field of modeling and simulation of regenerative systems.

Biological cells are not well-stirred environments, however, modelling and simulation methods based on this assumption are mostly better established theoretically and easier to apply than those that do not. In this project, we will establish concepts for the discrete, stochastic modelling of those processes taking the spatial dimensions of the system and actors into account. An important aspect will be the investigation of expressiveness. We will explore existing approaches to discrete, stochastic modelling and evaluate their suitability for describing continuous coordinates as well as lattice-based discretisations of space, and integrating various forms of paticle movement from Brownian motion to directed transport. Ultimately, we aim to link this work with the multi-level modelling approach developed in project Uhrm_A1.