Deciphering the information on genome sequences in terms of the biological function of the genes and proteins is a major challenge of the post-genomic era. Currently, the bulk of function assignments for newly sequenced genomes is performed using bioinformatics tools that infer the function of a gene on the basis of sequence similarity with other genes of known function. It is now well recognised that these primary, sequence similarity-based function annotation procedures are frequently inaccurate and error prone. Continuing to use them without clearly defining the limits of their applicability would lead to an unmanageable propagation of errors that could jeopardise progress in Biology. On the other hand, various novel bodies of data and resources are becoming available. These provide information on context-based aspects of the biological function of genes, namely on physical and functional interactions between genes and proteins, and on whole networks and processes. In parallel structural genomics efforts world wide are providing a much better coverage of the structural motifs adopted by proteins and on their interactions. The availability of these additional and novel data offers an unprecedented opportunity for the development of methods for incorporating higher-level functional features into the annotation pipeline.
The GeneFun project aims at addressing these two important issues. The issue of annotation errors will de addressed by developing criteria for evaluating the reliability of the annotations currently available in databases. These criteria will be used to assign reliability scores to these annotations and will be incorporated into standard annotation pipelines, for future use. The issue of incorporating higher-level features into functional annotations will be addressed by combining sequence and structure information in order to identify non-linear functional features (e.g. interaction sites), and by integrating available and newly developed methods for inferring function from higher-level and context-based information (protein domain architecture, protein-protein interaction, genomic context such as gene order etc.).
To achieve these aims several European groups with strong track record in developing novel methods and analyses in comparative genomics, structural- and systems- oriented bioinformatics, and in information technology, have teamed up with an experimental group from Canada, which is well known for its outstanding achievements in the field of structural and functional proteomics. The expected output of the GeneFun project is: improved procedures for inferring function on the basis of sequence similarity, a set of procedures for predicting non-linear functional features from sequence and 3D structure in a more automated way, and benchmarked procedures for predicting context-based functional features. Major efforts will be devoted to devising protocols that optimally combine the results from several methods. In particular Web-based servers to the individual and combined procedures will be developed, and made available to the scientific community. The community will be introduced to these new tools through open workshops and training sessions.