Our key work areas in research and teaching are methods for developing distributed systems and various applications that profit from being executed in a distributed/parallel computing environment. In particular, our research focuses on software infrastructures, services, and tools for cluster/grid/cloud computing, middleware for internet application environments, as well as approaches for the distributed/parallel solution of computation- and data-intensive problems from the areas of data analysis, simulation and optimization in scientific computing, computational engineering, and multimedia computing. Current application areas are biocomputing and image processing and analysis. Aspects of IT security, reliability, efficiency and scalability of infrastructures and applications are investigated as crosscutting topics.
In a joint project with the working groups Hüllermeier and Klebe, we investigate means for improving the computation times of methods for comparing spatial structures of possible protein binding sites. The main task is to determine which proteins can carry out mutually identical or similar functions, so that, for example, gene configurations of microorganisms can be reduced to a minimum as part of synthetic microbiology research. The SEGA algorithm developed by the working group Hüllermeier was efficiently parallelized, both by taking advantage of parallel computations on modern graphics hardware (GPGPUs), and by distributing computation to the nodes of a Cloud infrastructure. It was possible to achieve a running time of less than 2 milliseconds per binding pocket comparison. As a result, the pairwise comparison of 144.849 binding sites was performed in an acceptable time period of about three weeks. Through this collaboration, a full comparison of the binding pockets of high-resolution crystal structures that were previously recorded in the Protein Data Bank (PDB) was performed for the first time. Based on this data, a full comparative analysis can be done for the entire protein space, which not only allows a classification of the protein space into structurally and functionally similar, homologous and non-homologous protein groups, but also supports the systematic search for unexpected similarities and functional relationships.
Our joint project with the working groups Lenz and Mösch deals with modeling the dynamic behavior of Fus3/Kss1-MAPK signal pathway of the yeast Saccharomyces cerevisiae that controls at least two distinct differentiation programs, the fusion of sexual partner cells and biofilm formation. The current mathematical model describing the dynamics of the biofilm formation that are important for cell-cell interaction will be enhanced in oder to be able to investigate the dependency of binding properties and parameters such as strength of current, gravity, and buoyancy. The model should be designed to allow quantitative statements to be made that can be validated experimentally. Since the computational runtime required for the simulation of the model is very high on regular computers, the computations should be performed in parallel on multiple cores of a computer cluster (or in a cloud) and GPGPU platforms. This requires that the existing software, a variant of an "N-body simulation", has to be redesigned to consider the aspects of distribution, communication, and synchronization associated with parallelization and to take the algorithmic iterations and the particle motions in space into account.