ABSTRACT
The genome sequence database (GSDB) is a complete, publicly available relational
database of DNA sequences and annotation maintained by the National Center for
Genome Resources (NCGR) under a Cooperative Agreement with the US Department of
Energy (DOE).
GSDB provides direct, client-server access to the database for data contributions, community annotation
and SQL queries. The GSDB Annotator, a multi-platform graphic user interface, is freely available. Automatically
updated relational replicates of GSDB are also freely available.
The initiation of genome sequencing projects and the development of automated
DNA sequencers have changed the style and pace of DNA sequencing. A laboratory
with a single automated DNA sequencer can generate upwards of 50 000 bases of
raw sequence data per day in steady state, and this rate is expected to
increase by an order of magnitude within the next few years. Two complete
microbial genomes have now been completed (
1
-
2
), several more are well underway, and dozens are expected to be finished by the
end of the decade. The complete genome of
Saccharomyces cerevisiae
(
3
-
5
) is expected to be finished in 1996, and that of
Caenorhabditis elegans
(
6
-
7
) is expected to be complete by 1998. Over 80 million base pairs of human
expressed sequence tags (
8
) have been sequenced and analyzed (
9
). The era of efficient, inexpensive, high throughput DNA sequencing has
started. The genome sequence database (GSDB) is designed to meet the community
wide challenges of managing, interpreting, and using DNA sequence data at an
ever increasing rate.
High throughput, genome scale sequencing presents both technical and social
challenges. Sequence production laboratories employ a variety of strategies,
from high redundancy shotgunning producing highly accurate, completed sequences
to directed strategies such as Ordered Shotgun Sequencing (
10
), sequence mapped gaps (
11
) or transposon based walking (
12
) to sampling strategies that produce ordered sequence fragments separated by
gaps of known length (
13
). A laboratory may choose to release data to the public at any stage in the
production process with any of these strategies; public sequence databases
must, therefore, be prepared to manage sequence data in any form from complete,
multi megabase contigs to collections of sub kilobase fragments related by
order, orientation, and distance information. Moreover, multiple laboratories
will often be involved in the completion of a sequenced region: one laboratory
may generate mapped sequence samples, while many others produce highly accurate
sequences of subregions or of corresponding cDNAs from expressed genes. As a
region is further characterized, additional data showing diversity across
individuals, alternative expression products, or even gross rearrangements may
be added to the database. Finally, functional annotation of the sequence data
may be the work of tens to hundreds of individual investigators. Capturing the
latter data in a structured database is critical; too often, important
functional results are presented only in the printed literature, and are
therefore effectively lost.
GSDB is designed to meet the requirements for a community sequence database
outlined by Waterman
et al
. (
14
). GSDB supports complex,
ad hoc
queries in a standard language, SQL (
15
). GSDB represents sequence data produced by any strategy, and supports the
contribution of additional sequence data or structural or functional annotation
by multiple researchers. GSDB extends the Electronic Data Publishing paradigm (
16
) from a model in which the database is viewed as a primary publication for data
not appearing in the traditional literature to a model in which the database
serves as a multi user laboratory database for the entire molecular biology
community (
17
). Multiple sequences from a given region and structural and functional
annotations on sequences are viewed, in this model, as independent
observations, with authors and unique identifiers. GSDB provides multi user
editing capabilities, with the necessary authorship, data security, integrity
checking and versioning mechanisms needed to ensure that multiple authors do
not overwrite each other's work. A mechanism is also provided for individuals
or groups to define their own curated views of the data, which include whatever
sequences and features they select. Database users may choose to access the
entire database, a particular view maintained by an editorial group that
imposes particular standards, or a view based on data relevant to a particular
set of interests. GSDB functions, therefore, both as a community laboratory
database and as a collection of multiple, virtual, specialty databases.
GSDB version 2.2 replaces the notion of an `entry' employed by the archival
sequence databases with separate structures representing sequences and
annotation. Both sequences and annotations are assigned their own individual,
unique accession numbers that serve as permanent identifiers. This
architectural change reflects the importance of structural and functional
annotation as biological data, not merely notes about sequences. The addition
of individual identifiers to annotations simplifies retrieval. It also allows
multiple users to contribute annotation to a given sequence without updating an
entire entry record with each change.
Aligned sets of sequences are represented as a fundamental data type in GSDB
2.2. The alignment structure allows the representation of multiple fragments
from one clonal source (e.g., a single cosmid or bacterial artificial
chromosome), multiple isolates of a single region from different sources (e.g.,
multiple isolates of a virus or disease gene), and multiple cDNAs from a
genomic region (e.g., alternatively spliced mRNAs). Complete sequences are
maintained in the alignment, with both coordinate based and sequence difference
representations available for queries. The collection of covering sequences
from each aligned set provides, moreover, a truly non-redundant sample of the sequences in the database.
Discontiguous sequences obtained by genome sampling strategies, EST projects or
exon sequencing are maintained as sets of ordered, oriented fragments linked by
distance and uncertainty values. This structure can also be used to represent
sequence tagged site (STS) maps as discontiguous sequences spanning entire
chromosomes. Discontiguous sequences can be included in alignments; hence
regional or chromosomal sequences can be assembled by aligning multiple,
independently derived sequences to a discontiguous sequence scaffold.
GSDB 2.2 represents sequence confidence with a span oriented real confidence
value. This confidence value is qualified by flags indicating whether a
sequence is single or multiple pass and whether one or both strands have been
sequenced. Confidence values can, in the limit, be assigned to single bases.
The collection structure in GSDB 2.2 allows any set of sequences and annotations
to be grouped into a queryable view. Any individual or group can create
collections for their own or public use. Collections provide a mechanism for
individuals or groups to curate or referee both sequences and annotations; they
also provide a mechanism for assembling various non redundant or special
interest sets of data in structured form.
GSDB employs a three tiered open architecture, as shown in Figure
1
. The database is implemented with SYBASE, a commercial relational database
management system (RDBMS) (Sybase Inc., Emeryville, CA). An object-oriented layer of software written in C++ sits on top of the database.
This layer handles all database access and communication for a variety of
applications. The GSDB Annotator, a multi platform graphic user interface, is
the major system application provided as part of GSDB 2.2.
We have implemented a new schema structure to respond to the changing needs of
the genome community. Figure
2
provides a high level view of database element interrelationships,
incorporating the new data types discussed. A complete description of the
schema is available by anonymous ftp from ftp.ncgr.org.
GSDB accepts input data interactively via the GSDB Annotator, via direct
database-database transactions or via parsable files, and delivers output via the
GSDB Annotator, SQL queries and traditional flatfiles.
GSDB Annotator
. The GSDB Annotator is an interactive browser and editor client for GSDB. It
runs on multiple platforms including SUN, MacIntosh and PCs with Windows
systems (available mid 1996). The GSDB Annotator displays database sequences
and analysis results in a graphic format that reflects the way r ouse-driven fill-in forms allow easy addition of annotation to sequences.
GSDB's versioning of changes to sequences and annotation provides citability
analogous to that offered by scientific journals. In addition to citability,
versioning enables researchers access to any given dated data set in order to
perform repeated or more extensive analysis with a fixed set of data.
Versioning in GSDB is accomplished through a dual database design in which the
production database contains only the most recent versions of all data, and a
versioning database contains all historical data. All versioning is done at the
row level and each row that is inserted or updated in the production database
results in a new row being added to the versioning database. All rows in the
versioning database are time-stamped.
Design of the next version of GSDB, 3.0, is well underway. The major structural
change between GSDB 2.2 and GSDB 3.0 will be the capability of representing
features as shared across multiple sequences. For example, an exon feature may
be shared, in GSDB 3.0, across multiple genomic samples that show sequence
variants, and across multiple cDNA sequences that show alternative splicing.
Shared features will greatly enhance the queryability of GSDB.
The representation of aligned sets of sequences will be enhanced in GSDB 3.0 to
explicitly include relationship types, including substrate-product (e.g. genomic to mRNA) and instance-type (e.g. individual isolate to selected representative)
relations between sequences. Adding these relations will improve the
representation of both expression pathways and collections of isolates that
provide mutation or other diversity data. GSDB 3.0 will also include an
improved representation of binding relations between sequences.
As the richness of biological annotation grows, the ability to link to external
databases that provide additional detail or other types of data increases. GSDB
3.0 will provide capabilities for linking a variety of data to multiple
external databases in a way that will support complex inter database joins.
Development plans and draft specifications for GSDB 3.0 will be posted on
http://www.ncgr.org/gsdb as they become available. Comments on these documents
and requests for additional functionality are appreciated and will be
acknowledged.
DNA sequence data may have application in any area of biology, biomedicine,
agriculture or biotechnology, from basic research in molecular mechanisms,
developmental genetics or evolutionary systematics to epidemiology, vaccine
design, plant breeding, bioremediation or industrial process development.
Conversely, results obtained in any area of biological or biochemical research
may be relevant to the structural or functional characterization of a sequenced
region. No single database or collection of curators can capture all of
biology, or foresee the requirements of all potential applications of
biological knowledge. GSDB is, therefore, designed to function as one node in a
federation of autonomous, independently curated biological databases spanning
many domains and interest areas. As commercial quality tools for distributing
complex queries across the Internet become available, the sequence and
annotation data maintained in GSDB will become progressively more tightly
linked with relevant information from other parts of biology and biotechnology.
Information about GSDB as well as data contribution and access tools are
available on our World Wide Web site, http://www.ncgr.org, or contact us at one
of the addresses below.
GSDB WWW Site
http://www.ncgr.org/gsdb/gsdb.html
GSDB WebSub Submission Tool
http://www.ncgr.org./gsdb/WebSub.html
GSDB WWW on line updates
http://www.ncgr.org/gsdb/update.html
Assistance with WebSub
WebSub{at}gsdb.ncgr.org
GSDB Submissions
datasubs{at}gsdb.ncgr.org
GSDB Updates
update{at}gsdb.ncgr.org
Anonymous FTP
ftp.ncgr.org
Relational Replication Startup
dnr{at}ncgr.org
GSDB User & Public Accounts
gsdb{at}ncgr.org
The National Center for Genome Resources (NCGR) is a private, not for profit
[501c(
3
)] corporation established in 1994 to provide information and other resources
generated by the genome projects and related research and development to the
public and private sectors. NCGR's public projects include GSDB and a database
of information on the ethical, legal and social (ELSI) issues raised by genome
projects. Requests for further information about NCGR can be addressed to
ncgr@ncgr.org or to:
Coordinator of External Affairs
National Center for Genome Resources
1800 Old Pecos Trail
Santa Fe, NM 87505, USA
[1] 505 982-7840 (voice) 7690 (fax)
The Genome Sequence DataBase is supported by Cooperative Agreement 95ER62062
between the National Center for Genome Resources and the US Department of
Energy, Office of Health and Environmental Research. NCGR is also supported by
the US Small Business Administration under Award SB OT 94-001.
Web
. Our World Wide Web Server (http://www.ncgr.org) provides several different
access mechanisms to GSDB. Applications are available for adding new sequences,
reporting updates, and retrieving responses to internal as well as cross
database queries.
Accounts
. GSDB user accounts are available on request. All external users may also
access the public readonly account. All accounts allow interactive SQL access
from any machine on the Internet.
Data exchange with archival databases
. Each day GSDB exchanges data in the form of flatfiles with the European
Molecular Biology Laboratory, the DNA DataBank of Japan, and GenBank at the
National Center for Biotechnology Information. This exchange process enables
all sequence databases to maintain a complete and up to date collection of
nucleotide sequences.
Bulk data contributions
. Large sequencing centers make up the majority of bulk data contributors. To
facilitate the entry process of bulk sequence data, GSDB supports both file
based and direct SQL approaches. With the file based approach, users generate
custom files that are processed through the GSDB object layer into the
database. Direct SQL, however, allows users to download data from their local
database directly into GSDB via custom software.
REFERENCES
Return