Protein fold recognition

By the end of 1998 the sequencing of C.elegans has been completed, revealing the first complete genome of a multicellular organism, containing more than 19,000 predicted protein products. In a few years the number of protein sequences, will be in the millions. Interpretation of sequence data requires assignment of biological functions and a knowledge of the three dimensional structure of the protein products. Experimental determination of each structure is impossible, but a combined approach of experimental structure determination and computational approaches promises success. Currently structural genomics explores possibilities to determine a representative set of protein structures that covers all biological proteins. All other protein structures could then be modelled on demand from a related template. The number of distinct protein folds is expected to be in the order of 10,000. The goal seems achievable within five to ten years, but it is particularily important to avoid redundant structure determinations. Fold recognition is a recent technique that can be instrumental in this effort, and it is an asset in the modelling of structures from templates. The technique can be used to uncover structural relationships among proteins that are not recognizable on the sequence level. Using fold recognition tools redundant structure determinations can be reduced, saving valuable resources in the determination of a representative set of folds. Results obtained in the recent CASP experiment (critical assessment of techniques for protein structure prediction) confirm the applicability of fold recognition techniques in structural genomics. The goal of this project is to develop a benchmark for fold recognition and to use in improving alignment quality and fold assignment reliability and in particular in tuning the technique for high throughput applications and automated assignments. The optimized technique will be tested in CASP4, the public blind test scheduled for the year 2000 and will be applied to the C.elegans protein products. A database for the C.elegans genome will be implemented and maintained, with a particular emphasis on structural information derived from fold recognition augmented by functional and additional relevant information available on the web. The data base is intended to be an asset for structural genomics. The data base can be queried for proteins for which no structural information is available. These are prime targets for experimental structure determination. A major goal in molecular biology is the structural and functional characterization of all human gene products. In this endeavour computational approaches to structure determination are an essential component. In this respect the C.elegans project serves as a pilot study for the human genome.

The recent sequencing of genomes of complex organisms are lasting milestones in the biological sciences. In particular the sequencing of the human genome is seen as a major breakthrough that will yield to significant advances in medical research and human health. The result of genome sequencing is the linear sequence of the bases along the chromosomes. In principle this sequence contains the blue print of the organism. However, to understand its meaning this information needs to be decoded. The first step is the identification of genes, those regions along the DNA that code for proteins, and the characterization of these gene products. This process of `genome annotation` yields information about the function of the individual genes and proteins and their interactions in the cell. Genome annotation is a most important task of bioinformatics. Bioinformatics provides computer based tools which essentially expoit the enormous redundancy in the biological world. New sequences are compared to already known and annotated sequences stored in various data bases. Whenever a hit is found it is possible to infer the function of a gene or protein and in a number of cases it is even possible to construct a model of the three dimensional structure of the encoded protein. In the current project we developed an annotation system, which integrates an arsenal of established and new search tools, collects annotation data from various public data bases, and annotates the proteins of whole genomes in terms of function and structure in an automated and expert driven mode. The current version is accessible at www.came.sbg.ac.at and contains the annotated genomes of homo sapiens, mus musculus, caenorhabditis elegans and arabidopsis thaliania. The annotations contain properties of the protein sequences, structures, functions, and the localisation in the cell. A particular strength of this system is the integration of structure prediction techniques that have been developed in previous projects. These techniques provide models of the three dimensional structures of proteins, which is often the key to understand the function and biological role of the protein. On a global level, the current annotation platform identifies 29,000 human proteins that are related to a sequence of known annotation. These are 44% of the 65,000 of the currently known human genes, in other words almost half of the human gene products are characterized to some extent. On the other hand the coverage in terms of the whole sequences is only 27%, i.e. in many cases the hits cover only part of the sequences. The annotation system is used in collaborative projects to identify proteins which trigger the synthesis of antibodies after infection by a pathogen (in collaboration with Intercell, Vienna, and ProCeryon Biosciences, Salzburg). The experimental result are sequences (in the order of several hundred) whose functions and structures need to be identified with the goal to find therapeutic strategies that are suitable to interfere with the infecting agents. In this project the annotation engine and fold recognition are essential in identifying several target proteins that are amenable for drug development.