The Pathogen Sequencing Unit (PSU) at the Wellcome Trust Sanger Institute sequences a large number of diverse prokaryotic and eukaryotic genomes (http://www.sanger.ac.uk/Projects/). In recent years, new sequencing and assembly technologies and collaborations between sequencing institutes have dramatically increased both the output and quality of sequence data. Maintenance and dissemination of such data required the development of an integrated, publicly accessible database.

GeneDB stores sequence data and analyses generated via automated annotation pipelines, prior to manual annotation. The analysis pipelines include gene finding algorithms, protein feature predictions, BLAST and/or FASTA searches against nucleotide, protein and customized databases, protein domain and/or family search results and electronically inferred and manually revised gene ontology associations (GO) (1). Search results are continually reviewed during the curation process and complemented by additional datasets (Fig. ​(Fig.11).

The GeneDB homepage supplies links to the individual organism homepages. From these, researchers can take advantage of numerous ways to retrieve data and construct searches according to individual preferences and requirements. Clickable chromosome and contig maps, searchable text indices and browsable catalogues [GO assignments (1), descriptions, products, domains] provide fast and easy access. An additional query interface supports a wide range of queries on sequences and (curated) annotations stored in GUS, with the ability to combine searches with the Boolean operators AND and OR. For example, users can select all proteins of a specified length range with a specified number of introns. Other query options include GO assignments, keywords, chromosome, protein domains and predicted protein sequence features. The queries in each session are tracked via a history page, allowing further refinement of searches and downloading of results as a nucleotide or amino acid FASTA file. Furthermore, a variety of sequence similarity search facilities are available through GeneDB. In addition to WU-BLAST, GeneDB also supports omniBLAST, which permits searching across a set of selectable databases. An iterative BLAST (PSI-BLAST) search suited to the identification of distant homologues is envisaged to be available shortly. Peptide sequences can be searched with either user-specified motifs or using the peptide mass identification tool EMOWSE, part of the suit of EMBOSS open-source software tools (2). An alternative approach for accessing genes of interest is to use the official browser of the GO consortium, AmiGO. Several different methods are available for querying the data both externally (http://www.godatabase.org/) and internally via GeneDB, all of which include direct links to the gene pages.

Experienced biologists curate data for six of the organisms in GeneDB (Table ​(Table1).1). Such curation involves several aspects, all aiming to facilitate data querying and retrieval. First as a number of organisms are sequenced by more than one sequencing centre, it ensures consistent annotation across the whole of the respective genome. Secondly, sequences and their annotation are updated according to new submissions to public databases, publications and contributions by the wider scientific community (Fig. ​(Fig.1).1). Public information is used not only to verify existing gene models and annotations but also to add value, enabling users to retrieve groups of genes/proteins not possible by purely computational methodologies. Wherever possible, controlled vocabularies such as GO (1) are used. In the absence of such vocabularies, statements are captured in structured syntax either in the description lines (Fig. ​(Fig.2)2) or in a dedicated curation field, providing a concise summary of the major aspects of a gene’s biology. Links are provided to PubMed records and other resources used to compile the statements. Text indices point to other products sharing the same description line.

GeneDB code development will continue to concentrate on integrating GeneDB with the GUS schema. Part of this development is a collaboration with the GUS team at the Computational Biology and Informatics Laboratory (CBIL, University of Pennsylvania) to design a common web interface architecture, permitting the creation of customized web pages to suit individual database requirements [e.g. GeneDB, PlasmoDB (14), AllGenes (http://www.allgenes.org/)].