Previous and Current Research

Our group has long cultivated an interest in the fundamental algorithms for RNA secondary structure prediction. Bioinformatics tools that were born out of this interest in the 1990ies were the visualization tool RNAmovies and a heuristic method for riboswitch prediction, paRNAss. With the increased recognition of the manifold regulatory functions of RNA, bioinformatics topics related to RNA have begun to dominate the work of our group. Upto 2011, we have produced 14 RNA-related tools, available in open source for download, or for interactive or webservice use via the Bielefeld Bioinformatics Server. The program RNAhybrid (2004) is a widely appreciated tool for the prediction of micro-RNA targets. It combines comparative sequence analysis with the computation of hybridization energy based on thermodynamics, and careful statistics. It allows for, but does depend on heuristic restrictions imposed by many competing programs, such as the much debated “seed-hypothesis” for micro-RNA and mRNA interaction.

Classical RNA secondary structure prediction excludes the socalled pseudoknots for reasons of computational complexity, in spite of their physiological relevance. The programs pknotsRG (2004, 2007) and pKISS (2010) extend structure prediction to certain types of pseudoknots which occur frequently in functional structures.

The concept of “RNA shape abstraction” was introduced in 2004 and has found a variety of applications. Abstract shapes characterize RNA secondary structure by arrangement of structural components, irrespective of size and sequence content. They help to find significant near-optimal structures (RNAshapes, 2004, 2006), which opens a road to comparative structure prediction in the difficult case when structure is more preserved than sequence (RNAcast, 2006). Abstract shapes collected in a precomputed index can also significantly speedup the search of structural RNA family models (RNAsifter, 2008). These and other tools are currently used in a DFG project in cooperation with Anke Becker (formerly CeBiTec, now University of Freiburg), where we combine experimental and computational methods in a large scale search for noncoding RNA genes in S. meliloti and related Alphaproteobacteria.

Future Projects and Aims

Over the years, our group has worked out the method of “Algebraic Dynamic Programming” (2004), which has been used to great advantage in our own tool development projects. Our recent system “Bellman’s GAP” makes this technology available for others, and we seek cooperations and new applications to develop it further. On the theoretical side, a recent review (2011) has shown that there is a lot of unexplored ground with covariance models, designed to describe structural RNA families. Improvement of such models by enhancing them with alternative semantics will constitute a major part of our future work.