Genomic informatics : toward an integrated view of genomes
Symposiums organized by William M. Gelbart from October 18 – 22 and 22 – 26, 1997
Part I October 18 – 22, 1997
Ashburner Michael (University of Cambridge, Great Britain), Bayraktaroglu Leyla (Harvard University), Befos Panayiotis (EMBL Outstation, Cambridge, Great Britain), Chillemi Joseph (Harvard University), Crosby Madelin (Harvard University), Drysdale Rachel (University of Cambridge, Great Britain), Emmert David (Harvard University), Gelbart William (Harvard University), Gilbert Don (Indiana University, Bloomington), Grey Aubrey (de) (University of Cambridge, Great Britain), Kaufman Thomas C. (Indiana University, Bloomington), Lewis Suzanna (University of California), Matthews Beverley (Harvard University), Millburn Gillian (University of Cambridge, Great Britain), Rindone Wayne (Harvard University), Russo Twombly Susan (Harvard University), Tolstoshev Carolyn (National Institutes of Health, Bethesda), Whitfleld Eleanor J.(University of Cambridge, Great Britain).
Part II : October 22 – 26, 1997
Ashburner Michael, University of Cambridge, Great Britain, Bairoch Amos (Groupe Swiss-Prot, University of Geneva), Cherry Michael L. (Stanford University School of Medicine, USA), Danchin Antoine (Institut Pasteur, Paris), Eppig Janan (The Jackson Laboratory, Bar Harbor, USA), Gaasterland Terry (University of Chicago), Garrels James I. (Proteome, Inc., Beverley), Gelbart William (Harvard University), Goodman Nathan (The Jackson Laboratory, Bar Harbor, USA), Hodgkin Jonathan (Medical Research Center, Cambridge, Great Britain), Kafatos Fotis C. (European Molecular Biology Laboratory (EMBL), Heidelberg), Kaufman Thomas C. (Indiana University, Bloomington), Letovsky Stanley (John Hopkins Medical Institute), Lewis Suzanna (University of California), Matthews Kathleen A. (Indiana University), Polacco Mary (University of Missouri – Columbia), Riley Monica (Marine Biological Laboratory, Woods Fiole, USA), Silver Lee M. (Princeton University), Smith Temple (Boston University), Stoesser Guenter (EMBL Outstation, The European Bioinformatics Institute, Cambridge, USA), Thierry-Mieg Jean (Centre de recherches en biochimie macromoléculaire (CRBM) du CNRS, Montpellier), Zollner Alfred 166 (Max-Planck Institute of Biochemistry, Martinsried).
The Treilles conference on Genetic and Genomic databases was separated into two meetings organized by the principal investigators of FlyBase. The first of these meetings focused on specific issues of one such databasе — FlyBase — winch maintains and presents to the public the core genetic information on the fruit fly, Drosophila melanogaster and related flies. The purpose of the second meeting was to address the general issue of how different biological databases can be interconnected.
One of the true revolutions in biology over the fast fifteеn years is the revelation that most genes responsible for controlling cellular and developmental events are quite ancient, and descendants of these genes have similar cellular and developmental roles in evolutionarily diverse organisms. Thus, the modern biologist cannot be a specialist, simply working on his or her particular organism without reference to relevant information emerging from many different research systems. Rather, we will increasingly see that the boundaries between the traditional biological disciplines will blur and disappear as multi-disciplinary research programs become a requisite part of the scientific landscape.
The need for integration of information is made even more urgent by the progress of the large scale projects developed as part of the Human Genome Initiative. Tremendous amounts of sequence level and other genomic information are already emerging from these projects on the human genome and the genomes of model research organisms. The amounts of data emerging from these projects will only increase over the next several years. In addition, a great deal of important information continues to come from individual, focused research laboratories, and in fact it is the biological information from these individual laboratories that will put « flesh » on the skeletal information generated by the genome projects. The ability to integrate all of this information in a manner that allows researchers to effectively draw out the information that is of relevance to them is thus of major importance to the success of the world-wide genome effort. It was the goal of the conference to promote this effort by bringing together leaders of various genomic informаtics efforts to share viewpoints and visions on how to best provided this integrated view of genomic information to the international research community.
FlyBase : The Drosophila Genetic Database
Drosophila melanogaster is one of the most important animal models for genetic research, both historically and at present. Many genetic systems and biological insights have served to pioncer similar studies in many different animal systems, including vertebrates and humans. This conference took place at a very important time in the evolution of FlyBase. This was the first meeting of the project participants since it was agreed that the scope of FlyBase would greatly expand to include the data aid public informatics efforts of the two major Drosophila genome projects (BDGP — Berkeley Drosophila Genome Project, and EDGP — European Drosophila Genome Project). The goal of the new FlyBase consortium is to produce a seamlessly integrated, publicly available database containing all of the genetic and genomic information on Drosophila, both from the large scale genome projects and from the literature generated by the Drosophila scientific community. In addition to being the inaugural meeting of the new FlyBase collaboration, it was also a time when the next five year plan for FlyBase needed to be agreed upon. The specifics of the new collaboration and the five year plan were the subject of this conference.
Major issues involved the nature of the dataflow and division of responsibilities within the new FlyBase project, aid the structure and organization of the new FlyBase public WWW inter- face. It is through this WWW interface that most users access FlyBase data.
In brief, the BDGP and EDGP groups will be responsible for the annotation of the large scale sequences that they generate, whereas the Cambridge, Harvard and Indiana sites will continue to be responsible for the capture of genetic and genomic information from the literature. BDGP, EDGP and Harvard will jointly maintain and update sequence level annotation. Each of these three groups discussed in detail the tools and approaches thеу have developed for their sequence annotation responsibilities.
Database design and implementation
One of the complications of the FlyBase project (as well as one of its strengths) is that it involves several groups dispersed in the U.S. and Europe. Master databases for different subsets of FlyBase are maintained at different sites, according to the data capture responsibilities. It was agreed that for the short term, this distributed model would be maintained, with BDGP being responsible for maintaining the database of sequence level information (using the Informix database engin), Cambridge maintaining the master working database for bibliographic, genetic and database cross-reference information (using a flat-file structure), and Harvard maintaining an integration of the Cambridge data with Harvard’s curated molecular and transposon data (in the Sybase database engine).
In the longer term, FlyBase will explore the production of a single integrated database. The two likely candidates for this are Sybase and Informix, since there is already a considerable investment in schema and software development in these two platforms. One attractive feature of Sybase is its replication server software, that in principle allows multiple copies of a database to be maintained in synchrony. Harvard will be experimenting with the replication server technology during the coming year. The outcome of this experiment will affect further decisions about database engines for Cambridge and the integrated database.
The public FlуВasе interface
The long term goal is to have one FlyBase WWW structure mirrored at numerous sites around the world. It is likely that, in large part, this will mean that the web browser and querying tools will continue to utilize flat files derived from the central relational database(s). The general page layout is being redesigned, in part for visual effects and in part to accomodate the new classes of data in FlyBase. Extensive new graphical interfaces and querying systems are being written in Java, and will permit powerful and dynamic queries of map-based and anatomical based information.
Progress toward this long term goal will be incremental. In the near future, there will be a few new graphical interfaces pointing at fully integrated datasets, as well as numerous other interfaces and datasets that are not fully integrated. As more and more of the database migrates to an integrated structure, the independent interfaces and datasets will be retired and replaced by the fully integrated ones. It is anticipated that full integration will take about three years.
lnterconnectivity of Genomic Databases
One of the important lessons of the recombinant DNA era is that cellular and developmental pathways are often incredibly conserved over enormous evolutionary distances, and therefore that much of the « instruction manual » for building and maintaining diverse types of organisms are very similar. Further, a comparative approach to exploring and understanding the same pathways in different organisms is of great importance in understanding the underlying principles of biology. The public databases that house much of the data on the important genetic systems under study must reflect these relationships through a rich set of interconnections that will allow the scientific соmmunity to take advantage of the rapidly expanding universe of genetic information. This conference represented the first time that a focused group of community genetic database representatives met to consider issues of interconnectivity.
For the purposes of this conference, we were concerned with two general types of public databases :
1) sequence databanks, winch maintain information about speсifiс DNA, RNA or protein sequences. Typically, the data types are relatively restricted but al taxonomic groups are represented. Represented at the conference were the European Bioinformatics Institute (EBI), nucleic acid sequence databank (Guenther Stoesser) and the SwissProt protein databank (Amos Bairoch).
2) Community genetic databases, in which a much broader range of biological information is maintained about a much more restricted group of organisms, usually a single species, genus or family. Represented at the conference were leaders of databases and researchers of several of the most important genetic systems :
Escherichia coli – Monica Riley & Bacillus subtilis – Antoine Danchin
Comparative analysis of whole bacterial genomes –Terry Gaasterland
- The yeast, Saccharomyces cerevisiae : Alfred Zollner, J. Mkhael Cherry and James Garrels
- Plants: Zea mays (corn) – Mary Polacco & Arabidopsis thaliana – Michael Cherry
- Invertebrate animals: the nematode Caenorhabditis elegans – Jonathan Hodgkin and Jean Thierry-Mieg, & the fruitfly, Drosophila – William Gelbart, Michael Ashburner, Thomas Kaufman, Kathleen Matthews and Suzanna Lewis
- The malaria vector mosquito, Anopheles – Fotis Kafatos
- Vertebrate animals: The mouse – Janan Eppig and Lee Silver, & Humans – Stanley Letovsky
In addition, because of their general expertise in issues of genomic databases, present as general commentators were Nathan Goodman and Temple Smith.
The first section of the conference consisted of presentations of the individual databases, with an emphasis on the nature of the information that was being captured and the problems that were being encountered.
There is considerable heterogeneity in the level of information available in community databases of these different systems. The most detailed information is available for those microorganisms that have been subjected to intensive genetic analysis and whose genomes have been fully sequenced (bacteria, especially E. coli and B. subtilis, and yeast, S. cerevisiae). The C. elegans ACeDB includes extensive genetic, physical mapping and sequence level information, with the genomic sequence scheduled to be completed in 1998. A merger of genetic and sequence level information into one Drosophila database is underway. For other higher eukaryotes, the соmmunity database emphasis is much more on genetic and physical mар level information.
There is also considerable heterogeneity in the implementations of these varions databases. Several use relational database engines, such as Sybase, Oracle or Informix. Others are based on ACeDB software. Still others depend upon flat file formats. While details of the database schemas were not discussed, it is clear that the schemas have been developed independently and reflect the needs of the particular community database.
Finally, at the level of nomenclature, that has evolved according to the history and the viewpoint of the particular organism group.
In the second part of the conference, the focus of the discussions involved ways in which the biological databases might better interconnect.
One issue that received considerable attention was the establishment of a vocabulary to adequately describe the function of gene products (proteins and RNAs). Currently, there is no robust system for describing functions not associated with enzymatic activities, and even in the enzyme domain, there are aspects that are not well covered. Further, there are aspects to a functional description that do not relate dirеctlу to the activity of the protein, but rather relate to the process in which it participates. Finally, a system needs to be established that has sufficient breadth to encompass the very different descriptive needs of all phylogenetic groups.
The outcome of the discussion was to agree that a system that provided for many different ways to describe the function of gene product was needed. It was agreed that if there was too much depth to these hierarchies, then the ability to define vocabularies that could be used to establish broad interconnections would likely fall apart. A working group was established that would have the goal of identifying a few categories of vocabularies that would be useful, in which terms would be listed in a shallow hierarchy, of perhaps two levels. One type of category might be molecular functions, and another might be biological processes. The working group will report back to the meeting participants by email and continue a discussion that will hope- fully lead to the adoption of a common set of functional voca- bularies. Such a system will form a very important basis for direct interconnections between the community databases.
Another issue that was discussed was the heterogeneity of database engines and structures within the community databases. This issue proved very controversial. Some participants argued that much of the data housed in the individual organism databases was of a similar nature, and that the needs of the broader research community would be best served by housing these data in a common database. This would have several advantages, including the elimination of a duplicative effort to build and maintain these various database structures, and the ability to mine the database in a much more robust manner. Equally passionately, others viewed such a merged database as an unacceptable monolith, with one major concern being the necessity of forcing heterogeneous data types into a common but inadequate data model. There was no consensus on this issue, but it seems likely that a prudent approach would be to continuously evaluate which data classes are sufficiently well-understood that thеу can be placed in a merged, species-independent database. In this approach, the relationship between species-specific community databases and central databases will be continuously evolving.
The other approach to this issue is to use a common language to describe the kinds of objects that exist in each database. CORBA is one such standard that is under active development. The goal of such a standard is to have database definitions that would allow interconnections and querying tools that can simultaneously interrogate multiple free-standing databases. As a limited experiment, a working group headed by Jean Thierry- Mieg would attempt to make a CORBA definition of several of the ACeDB-based databases that have been developed. Again, this working group would report back to the participants by email concerning the outcome of this experiment.
The final issue that was discussed was that of genetic nomenclature. Because of the history of genetics in different systems, in which genes and mutations are often identified long before there is anу understanding of the molecular or biological roles of the gene product, there is no common agreement on how to name genes and alleles. Indeed, genetic nomenclature of different organisms operates by quite different conventions and is сеrtainlу a bander for researchers trying to understand related biology in different genetic systems. This issue is exacerbated by the problem of adherence to the estaЫished nomenclature of a given species. Thus, the scientific literature often refers to a given gene by several different names ; this is often for legitimate historical reasons, but it is very unhelpful to researchers trying to wade through this literature. The participants agreed that all community databases should maintain a list of valid gene symbols, gene names and gene identifier numbers, as well as synonyms and histories, perhaps maintained in a common database. Further, the group should approach the major biological journals to get their agreement to require the incorporation of these valid symbols, names and identifiers in al publications — essentially as nucleic acid sequence accession numbers are required by most journals.
This would entail considerable overhead to the community databases, in terms of establishing prospective assignment of gene symbols and names (some organism groups already do so), and in terms of maintaining a service for reviewing use of genetic nomenclature. The advantage however, would be a much cleaner and easily traversed literature, and less retrofitting of terminology during the annotation of the literature. A working group was established to put together a more specific proрosal with which to approach the journals.
In summary, this conference was an important beginning to consideration of several means by which more robust interconnectivity can be established. The approach that was adopted was to try to identify several short term, practical issues to address rather than to try to solve all problems at once. We appreciate the willingness of Fondation des Treilles to serve as the venue for the first in what will hopefully be a long series of discussions on how to bring together the genetic and genomic information in different systems.
- First meeting
Michael Ashburner – Describing gene function in genetic databases
Leyla Bayraktaroglu – Molecular level curation at FlyBase
Joseph Chillemi – FlyBase computer support for molecular curation
Мadeline Crosby – Molecular level curation at FlyBase
Rachel Drysdale – Genetic and bibliographic data in FlyBase
David Emmert – FlyBase molecular schema
William Gelbart – FlyBase project structure
Don Gilbert – Drosophila genome database client-server improvements, problems and plans
Aubrey de Grey – Genetic and bibliographic data in FlyBase
Suzanna Lewis – Database design for the FlyBase-BDGP-EDGP integration
Beverley Matthews – Molecular level curation at FlyBase
Kathleen Matthews – FlyBase public server development
Gillian Millburn – Genetic and bibliographic data in FlyBase
Wayne Rindone – FlyBase database design
Susan Twombly Russo – FlyBase computer database
Carolyn Tolstoshev – FlyBase database – compatibility issues
Eleanor J. Whitfiеld – Genetic and bibliograpbic data in FlyBase
- Second meeting
Michael Ashburner – Controlled vocabularies for cross-database interconnections
Amos Bairoch – Sequences, sequences, sequences ! Should databases be autоcracies or dictatorships ?
Michael J. Cherry – Saccharomyces genome database
Antoine Danchin – Bacillus subtilis, Eochenidia coli databases – Imagen (knowledge base) – The concept of neighbourhood for inductive approach of genomes
Janan Eppig – Mouse genome database and gene expression database
Terry Gaasterland – Automated genome annotation updates
James I. Garrels – The yeast proteome database (YPD)
William Gelbart – The FlyBase consortium
Jonathan Hodgkin – The nematode C. elegans : biology and databases
Fotis C. Kafatos – The wonderful world of non-model organisms
Stanley Letovsky – Future directions for public bioinformatics databases
Mary Polacco – Maize DB links
Monica Riley – Multidimensional classification of fonctions
Silver Lee M. – Mouse genetics-genomics, development, complex traits
Temple Smith – Database inconsistancies related to multidowsin proteins
Guenter Stoesser – International sequence DB collaboration
Jean Thierry-Mieg – ACEDE/JADE
Alfred Zollner – Yeast genomic database