Current Projects and web services

Protein structure comparison

 The description of similarities between the spatial structures of biological macromolecules is essential for the understanding of their function and evolution.  A focus of our research group is the development of bioinformatics tools which allow the rapid detection and display of these structural similarities. Since proteins often fulfill their biological role by interacting with other proteins (as protein complexes), our main interest is in structural similarities between macromolecular complexes.  The comparison of entire protein complexes often reveals interesting structural relationships which are hidden when comparing individual protein chains.
Strongly related to a pairwise comparison of protein structures is protein classification, where proteins are grouped into families according to mutual similarities.  Structure-based classification allows uncovering relationships which are not detectable by sequence-based comparison alone and thereby can provide new insights into protein evolution.
The tools developed for these projects are available as web services for pairwise structure comparison ( TopMatch) and fast structure database searches ( TopSearch).



Structure-based prediction of protein stability

MAESTROThe stability of proteins plays a major role in natural biological processes. A change in the stability through mutations can have serious consequences for such processes and may lead to severe diseases. In addition, stability is an important aspect in the design of proteins used therapeutically or biotechnologically. We developed  MAESTRO, a software tool for the fast and reliable prediction of stability upon point mutations. MAESTRO operates on three-dimensional protein structures and successfully employs statistical scoring functions as well as an ensemble of machine learning methods to fulfill this purpose.

Structure-based predictions of MHC-II binding peptides

MHCII3DMHC-II proteins (in human primarily HLA-DR, HLA-DQ or HLA-DP) bind peptides of different sizes and present these peptides on the surface of different cells of the immune system. Such an MHC-II/peptide complex is then recognized by T-cells, which triggers defense mechanisms. Besides the natural role of MHC-II/peptide complexes in defense of pathogens, they also play an important role in allergy, autoimmune diseases, and cancer. We developed the computational method  MHCII3D for the reliable prediction of potential binding peptides for the multitude of viable MHC-II variants.


Finished projects and web services

(services are not further developed and the databases are possibly not up to date)

Prediction and analysis of transcription factor binding sites on the DNA

 Transcription factors (TFs) are responsible for enabling or disabling the transcription of genes, and thereby determine which biomolecules (primarily proteins and functional RNAs) are produced, which subsequently define the different biological processes. Therefore, these processes may be perturbed by changes in the TF production or the loss of TF binding sites (TFBS) by mutations on the DNA. This malfunction is often caused by cancer.
We developed  D-Light on Promoters, a client-server based genome browser for organizing and visualizing sequence annotations, with a strong focus on transcription factor binding sites (TFBSs). An integrated scanner for predicting TFBSs, based on positional frequency matrices, enables the analysis of genomic regions, even if no or only little experimentally determined data are available.

Protein function score

 Public databases such as UniProt or NCBI protein provide information on numerous proteins. This information is of varying extent and reliability. The  protein function score (PFS) uses various fields of entries in the UniProt/SwissProt database in order to assess how funded this information is in terms of the function of the respective protein.

Reference structural alignments

 The comparison of the 3D structure of two proteins is very useful, e.g. to recognize evolutionary relationship when sequence comparison is no longer applicable. Structure comparison often involves the assignment of the corresponding homologous sequence positions. Creating this alignment can be difficult and thus error-prone. In order to compare different methods for 3D structure alignment with respect to alignment accuracy and their ability to recognize complex relationships like permutations or flexibility, we manually  compiled suitable data sets.


  Web service and a  stand-alone program for protein structure analysis. ProSa supports and guides your studies aimed at the determination of a protein’s native fold. It is helpful for experimental structure determinations and modeling studies.




  NQ-Flipper validates and corrects asparagine and glutamine side-chain amide rotamers in protein structures solved by X-ray crystallography.




 The decomposition of proteins into structurally distinct units is a common tool for classification and examination of protein structures.  TopDomain-Web provides an easy-to-use albeit powerful interface to a suite of techniques for the decomposition of protein structures.




  RefDens relates electron densities computed from a structural model to densities expected from prior observations on identical or closely related molecular environments. Strong deviations of computed from expected densities reveal unrealistic molecular structures. The electron densities solely depend on the molecular model, so that structure analysis and error detection are independent and hence complementary to experimental data.



  Canonical distributions of electrons can be used in the refinement of structural models obtained from X-ray, NMR, cryo-EM, and other imaging techniques, as well as structure prediction and molecular modeling. Difference density maps for structural models of any origin can be computed and downloaded. In addition, hybrid density maps are provided that combine canonical electron density maps with experimentally derived densities to aid in the real space refinement of crystal structures.



 Most common approaches to extract structural information from NMR chemical shifts are based on chemical shifts comparison. Therefore, consistent chemical shift referencing is required, as two chemical shift sets calculated using different reference compounds or referencing methods may not be compared in a meaningful way.  CheckShift reliably re-references uploaded chemical shift data to conform with the IUPAC standard.



 The  COPS-benchmark provides a convenient and easy to use standard reference for the analysis of sequence alignment methods, homology search tools and fold recognition techniques. Any method can be immediately related to several popular methods (BLAST, PSI-BLAST, FASTA, SSearch, COMPASS, and HHsearch).



 Signal peptide prediction by sequence alignment using BLAST. Predictions returned by  Signal-BLAST are transparent and easy to analyze.