Background and state of the art
The determination of the three-dimensional structures of proteins and nucleic acids as well as their complexes is a prerequisite for understanding their function. The bottlenecks of X-ray crystallography are (i), the crystallisation of the molecules of interest and (ii), the phasing of the X-ray diffraction amplitudes obtained from the crystals. This so-called phase problem requires the usage of heavy atom substitution of the native crystals. So far, direct determination of the phases of the X-ray reflections from the complicated relationships between the amplitudes has remained a dream for all but the smallest macromolecules.

The Project
We have developed a new computational approach that makes use of a genetic algorithm to solve the phase problem. This method has been applied to phase diffraction data of two medium-size protein structures. It remains to be shown that its applicability is general and not limited to extremes of resolution, symmetry, or molecular size. In order to improve the algorithm, we plan to incorporate dynamic fitness landscapes.

• Webster, G. and Hilgenfeld, R.: An evolutionary computational approach to the phase problem in macromolecular X-ray crystallography. Acta Cryst. A 57:351-358, 2001.
• Wilke, C.O. and Martinetz, T.: Adaptive walks on time-dependent fitness landscapes. Physical Rev. E 60:2154-2159, 1999.

Background and state of the art
The infectivity of several obligate or facultative intracellular bacteria such as Chlamydia pneumoniae, Chlamydia trachomatis, Legionella pneumophila, or Rickettsia depends on a protein named macrophage-infectivity potentiator (Mip). This protein has been shown to possess peptidyl-prolyl-cis/trans-isomerase (PPIase) activity, the substrate of which is unknown. While the bacteria mentioned above contain a homodimeric Mip protein, the version present in the eukaryotic parasite, Trypanosoma cruzi, is a monomer. The crystal structure of the Mip protein from Legionella pneumophila, the etiological agent of Legionnaires' disease, has been determined at 2.4Å resolution. A second crystal form diffracting to 1.8Å gave further insight into the dynamics of Mip. It has also been shown that Mip is inhibited by the PPIase inhibitor FK506 and the structure of the complex has been determined.

The Project
Three-dimensional models for the homologous Mip proteins from Chlamydia spp. will be constructed by homology-model building. Along with the X-ray structure of Legionella Mip, these will be used in virtual screening for and docking of potential new inhibitors. Hits found in this procedure will be synthesised in-house and tested for PPIase inhibition in vitro and for antibacterial effects in vivo.

• Riboldi-Tunnicliffe, A., König, B., Jessen, S., Weiss, M.S-, Rahfeld, J., Hacker, J., Fischer, G. and Hilgenfeld, R.: Crystal structure of Mip, a prolylisomerase from Legionella pneumophila. Nature Struct. Biol. 8:779-783, 2001.
• Hillisch, A., Pineda, L.F. and Hilgenfeld, R.: Utility of homology models in the drug discovery process. Drug Discov. Today 9:659-669, 2004.
• Hogg, T. and Hilgenfeld, R.: Protein crystallography in drug discovery. In: Comprehensive Medicinal Chemistry II, Vol. 3 (eds: Kubinyi, H., Triggle, D., Taylor, J.), Elsevier, Amsterdam, pp. 875-900, 2007.

Background and state of the art
The vast majority of eukaryotic proteins whose topogenesis starts at the endoplasmic reticulum are translocated across the ER membrane in a co-translational manner. However, there is clear indication that in all eukaryotic organisms some proteins exist, that despite being larger than 12 kDa are translocated only after completion of their synthesis. While some (but not all) of these proteins have been identified in the yeast Saccharomyces cerevisisae, none is known from other eukaryotic organisms. There is experimental evidence, that in yeast the major trait directing proteins of that size either into the co- or into the posttranslational pathway is coded in the signal sequence.

The Project
This collaboration has identified putative signals on yeast protein sequences that sign statistically responsible for the translocation pathway. By an iterative cycle of verification of the predicted signals using in vivo and in vitro translocation systems and bioinformatics analysis of the obtained experimental data we will first improve the prediction algorithm. Then this improved algorithm will be used for screening different eukaryotic genomes for potential posttranslational substrates. The correctness of this prediction will again be verified analysing selected substrates of different organismal origin in heterologous or homologous experimental systems. An improved model for this sorting process has potential practical applications i.e. in gene therapies and the understanding of pathogen physiology as it allows the assignment of subpathways of topogenesis to the different secreted proteins.

• Kim, J.T., Gewehr, J., and Martinetz, T.: Binding matrix: A novel approach for binding site recognition. J. Bioinform. Comput. Biol. 2:289-307, 2004.
• Meyer, H.-A., Grau, H., Kraft, R., Kostka, S., Prehn, S., Kalies, K.-U. and Hartmann, E.: Mammalian Sec61 is associated with Sec62 and Sec63. J. Biol. Chem. 275:14550 14557, 2000.

Background and state of the art
The control of gene expression is of central importance in biomedicine and molecular therapy. It can be facilitated via gene interruption tools which include small regulatory nucleic acids such as antisense oligonucleotides (asON), microRNA (miRNA), and small interfering RNA (siRNA). All of those tools are short oligonucleotides. Among the large number of possible species against a given target RNA, only a small fraction shows effective suppression of its target gene in living cells. The biological activity of all of those tools is influenced by local characteristics of the target RNA including local RNA folding. Experimental approaches to identify promising local target motifs and, hence, complementary gene interruption tools, have provided tools to define favourable sequences. However, experimental protocols share a number of crucial disadvantages. As insights into the structure-function relationship of asON as well as the role of sequence motifs increase, it becomes feasible to consider computer-based theoretical approaches for the design of biologically effective asON, miRNA, and siRNA.

The Project
We propose to develop computer-based tools for the automated theoretical design of gene interruption and to implement suitable algorithms. The development will be based on novel machine learning methods for motif detection and classification.

• Patzel, V. and Sczakiel, G.: Theoretical design of antisense RNA structures substantially improves annealing kinetics and efficacy in human cells, Nature Biotechnol., 16:64-68, 1998.
• Kim, J.T., Gewehr, J. and Martinetz, T.: Binding matrix: A novel approach for binding site recognition. J. Bioinform. Comput. Biol. 2:289-307, 2004.

Background and state of the art
For the investigation of the molecular basis of myocardial infarction the group of Heribert Schunkert and Jeanette Erdmann (head of the molecular genetics working group at the MKII) has access to genome-wide 500K Affymetrix SNP-Chip data both from 1,000 myocardial infarction patients with strong genetic background (index patients with at least one affected first-degree relative before the age of 60 years) and 1,700 population-based controls (funded by the EU commission through the sixth FP under reference LSHM-CT-2006-037593; Cardiogenics). The data-analysis is ongoing and will involve major bioinformatics investigations over the next four years. Preliminary analysis using PLINK software (Whole genome association analysis toolset) revealed several interesting chromosomal regions harbouring candidate genes for the pathogenesis of myocardial infarction. Information on sequence variations in genes and regulatory elements in regions tagged by the top ranked and replicated SNPs will then be enriched by a process of re-sequencing of exons and introns and of regulatory regions (identified by comparative genomics) in 96 DNA samples with the goal to identify the true causal variant. When genome variants are identified in genomic DNA, especially during routine analysis of disease-associated genes, their functional implications might not be immediately evident. Distinguishing between a genomic variant that changes the phenotype and one that does not is a difficult task. However, increasing evidence indicates that genomic variants in both coding and non-coding sequences can have unexpected deleterious effects on the gene transcript.

The Project
Current sequencing projects concentrate on functional mutations leading to amino acid substitutions or on variants located in promoter regions. Genomic variations with no apparent effect ("neutral polymorphisms") may have a significant effect on splicing (Pagani et al. Nature Reviews Genetics 2004) or timing of co-translational folding in case of synonymous mutations in exonic regions. Recent work by Kimchi-Sarfaty et al. (Science, 2006) shows that a synonymous SNP in the MDR1 gene results in a protein with altered drug and inhibitor interactions. They hypothesise that the presence of a rare codon, marked by the synonymous polymorphism, affects the timing of co-translational folding and insertion of the MDR1 protein into the membrane, thereby altering the structure of substrate and inhibitor interaction sites. In fact, this is the first report of synonymous mutation resulting in a pathogenic condition in humans.
The effect of these types of mutations is very difficult to spot, unless a functional assay is undertaken, which is in any case very difficult and time-consuming. Therefore, it is of particular interest to determine with in-silico methods those regions that may be influencing the splicing of the transcript or affects the timing of co-translational folding by using rare codons.

• Broeckel, U., Hengstenberg, C., Mayer, B., Holmer, S., Martin, L.J., Comuzzie, A.G., Blangero, J., Nurnberg, P., Reis, A., Riegger, G.A., Jacob, H.J. and Schunkert, H.: A comprehensive linkage analysis for myocardial infarction and its related risk factors. Nature Genet. 30:210-214, 2000.
• Fischer. M, Broeckel, U., Holmer, S., Baessler, A., Hengstenberg, C., Mayer, B., Erdmann, J., Klein, G., Riegger, G. and Jacob, H.J., Schunkert, H.: Distinct heritable patterns of angiographic coronary artery disease in families with myocardial infarction. Circulation. 111:855-862, 2005.

Background and state of the art
Adult stem cells play a key role in tissue homeostasis and regeneration. During these processes they are well regulated by an intricate interplay of extrinsic and intrinsic factors. When these stem cells are taken out of their natural environment and subjected to cell culture conditions, they show a complex, yet poorly understood behaviour. In particular they may spontaneously form three dimensional structures, while at the same time giving rise to different kinds of specialized cells. A detailed knowledge of the dynamics and microkinetics of these processes can lead to first mechanistic models for tissue formation and regeneration.

One promising approach to follow the complex in vitro behaviour is to do long term microscopical observations (time lapse microscopy), possibly in three dimensions. Using computer vision and machine learning techniques one can then identify and measure the relevant parameters.

The Project
In this project, time-lapse imaging of living cells is combined with computer vision and machine learning technologies to discover patterns in the in vitro behaviour of the cells. This non-invasive approach has the advantage that the cells are left undisturbed, since already small perturbations may alter the growth and differentiation characteristics of the cells. Moreover, the method allows for the investigation of large cell numbers, thus ensuring the statistical relevance of the obtained data.

The time-lapse data will be evaluated via advanced computer vision methods to extract several parameters like cell dimensions, migration speed, migration modus (individual or in droves), cell division rate, lineage tracking and combinations thereof. When combined with fluorescent markers, which report on specific cell properties, this approach can be a powerful method in understanding the complex in vitro behaviour of adult stem cells. The here gained expertize in analyzing complex biological data with advanced computational methods can be beneficial for other application domains.

• Kruse, C., Willkomm, D., Gebken, J., Schuh, A., Stoßberg, H., Vollbrandt, T. and Müller, P.K.: The multi-KH-protein vigilin associates with free and membrane bound ribosomes. Cell. Mol. Life Sci. 60:2219-2228, 2003.
• Aach, T., Mota, C., Stuke, I., Mühlich, M. and Barth, E.: Analysis of superimposed oriented patterns. IEEE Transactions on Image Processing 15:3690-3700, 2006.
• Kruse, C., Kajahn, J., Petschnik, A.E., Maaß, A., Klink, E., Rapoport, D.H., and Wedel, T.: Adult pancreatic stem/progenitor cells spontaneously differentiate in vitro into multiple cell lineages and form teratoma-like structures. Annals of Anatomy, 188:503-517, 2006.