Sequence Analysis Methods for Protein Function Prediction

The event of complete genome sequencing has opened new possibilities and promises. The pure availability of genomic sequences is of little help in life science research, albeit the development of techniques for sequence analysis did not keep pace with the progress in literal sequence information production. Whereas the delineation of protein sequences possibly encoded in the given genome data is sometimes difficult but generally possible, the insufficient understanding of biological function for proteins known only as conceptual translations has become the major bottleneck. With this project, we promise the solution of three important tasks with general impact for protein sequence analysis: Part 1: Low complexity regions (LCRs) are generally not well characterized in terms of function and structure. This might improve if a BLAST-like database search tool becomes available clustering LCRs with some type of similarity criterion into families. We want to develop such tool. This task cannot be solved with the recently applied apparatus for homology searches since the statistical criteria applied do not work for segments with strongly biased amino acid composition. Due to the degeneracy of the sequence within LCRs, it appears that the actual sequence of residues is very often not the critical point in their biological function compared with certain integral sequence characteristics (for example, their composition). We want to create a tool selecting families of sequence segments defined by integral sequence similarity criteria and to apply this method for LCR classification and functional characterization. Part 2: A large variety of biological features (structures and functions) such as many posttranslational modifications of amino acid residues in proteins or proteolytic scissions of polypeptide chains in cellular processes cannot be predicted with theoretical methods at all. Thus, large classes of sequences will remain non-annotated or incompletely annotated since the scope of the existing prediction techniques is limited. We want to develop techniques for the recognition of various lipid anchor sites in protein sequences (for example GPI-lipid anchors in non-animal taxa, myristoyl, farnesyl and palmitoyl anchors) as well as for cell cycle-specific protein cleavage points (for example, in the substrate proteins for separins in the early anaphase). If our resources and the research progress allow, we will extent the effort to other types of posttranslational modifications. Part 3: The existing methods (algorithms and software solutions) are not prepared for the generation of annotation of large amounts of sequences. Especially, automated implications (deduction heuristics) based on the combined output results of different prediction methods are impossible. We want to create a portable software solution ("Automated Sequence Analyzer") that can invoke different prediction methods, parse their output, and store this data in an electronically retrievable form. Specifically within this project, we want to develop the deduction algorithms that should mimic the routine activities of bioinformatics researchers if they invoke prediction programs and analyse their outputs manually. This effort has great importance in context with the functional annotation of sequences encoded on DNA chips.