Download Biological Ontologies in Rice Databases. An Introduction to the

Plant Cell Physiol. 46(1): 63–68 (2005)
doi:10.1093/pcp/pci505, available online at www.pcp.oupjournals.org
JSPP © 2005
Mini Review
Biological Ontologies in Rice Databases. An Introduction to the Activities in
Gramene and Oryzabase
Yukiko Yamazaki 1, 3 and Pankaj Jaiswal 2
1
2
Center for Genetic Resource Information, National Institute of Genetics, Mishima, Shizuoka, 411-8540 Japan
Department of Plant Breeding, Cornell University, Ithaca, NY 14853, U.S.A.
;
gene ontology (GO) (Gene Ontology Consortium 2004) [see
Appendix 1 (1)], for example, is one of the most successful
biological ontologies. Most genome databases of model organisms as well as DNA/protein sequence databases such as
SWISS-PROT/TrEMBL/Ensembl now share GO-IDs, identifiers for each term in the gene ontology. When several genes
from a variety of organisms associate with a GO-ID, a set of
genes associated with the GO-ID can be acquired from the GO
database without a sequence similarity search. Furthermore, a
set of genes associated with both higher and lower hierarchy of
the ID is also retrievable. Without using GO-IDs, it takes a
long time to obtain the same results, because of the problem of
genes sharing the same function having different annotation
(gene names or functional descriptions).
Although the term ‘ontology’ is not familiar to biologists,
the Enzyme Commission (EC) number and the taxonomic hierarchies, for example, can be regarded as pioneering ontologies.
Before the GO started in 2000, FlyBase (FlyBase Consortium
2003) [see Appendix 1 (2)] had started a program called ‘Links
to databases of other genomes’ in 1998, in which each Drosophila gene has a lineage of the homologous genes in other
organisms, e.g. ECOGENE:EG10396|glpF|FBgn0000180|bib.
At that time, the MENDEL Database (Price and Reardon
2001) provided the ‘Designations and families of sequenced
genes primarily in plants’. These trials and others have been
unified and have grown into the GO that covers all organisms.
In addition to GO, various types of scientific knowledge
have being structuralized under the name of ontology. While
conventional classifications have limitations concerning traditional form, definition and logic, ontology is more flexible.
Perhaps it would be more helpful to biologists to show examples instead of a general introduction to ontology. We will
review the current status of biological ontologies in Gramene
(Jaiswal et al. 2002, Ware et al. 2002) [see Appendix 1 (3)] and
Oryzabase (Yamazaki et al. 2003) [see Appendix 1 (4)]; two
comprehensive rice databases.
All URL addresses cited in this review are listed in
Appendix 1.
An enormous amount of information and materials in
the field of biology has been accumulating, such as nucleotide and amino acid sequences, gene and protein functions,
mutants and their phenotypes, and literature references,
produced by the rapid development in this field. Effective
use of the information may strongly promote biological
studies, and may lead to many important findings. It is,
however, time-consuming and laborious for individual
researchers to collect information from individual original
sites and to rearrange it for their own purpose. A concept,
ontology, has been introduced in biology to support and
encourage researchers to share and reuse information
among biological databases. Ontology has a glossary,
named dynamic controlled vocabulary, in which relationships between terms are defined. Since each term is strictly
defined and identified with an ID number, a set of data represented in biological ontology is easily accessible to automated information processing, even if the data sets are
across several databases and/or different organisms. In this
mini-review, we introduce activities in Gramene and
Oryzabase, which provide biological ontologies for Oryza
sativa (rice).
Keywords: Database — Gramene — Ontology — Oryzabase
— Oryza sativa (rice).
Abbreviations: GO, gene ontology; GRO, growth stage ontology;
PO, plant ontology; QTL, quantitative trait locus; TO, trait ontology.
Introduction
The word ‘ontology’ has originated historically from a
term in philosophy, which means ‘subject of existence’. It has
been used in the field of artificial intelligence as an explicit
specification of a conceptualization, and bioinformaticians
started to use this concept about 5 years ago. The purpose of
biological ontology is to share and reuse knowledge with
researchers, especially those studying different organisms. The
3
Corresponding author: E-mail, [email protected]; Fax, +81-55-981-6886.
63
64
Gramene and Oryzabase
Ontologies in Gramene
Gramene is a comparative genome mapping database for
grasses, using the rice genome as an anchor. Both automatic
and manual data curation are performed to combine and interrelate information on genomic and cDNA sequences, proteins,
various maps (genetic, physical and molecular marker map),
mutant phenotypes, quantitative trait loci (QTLs) and publications. As an information resource, the purpose of Gramene is to
provide added value data from public databases to facilitate
researchers’ ability to leverage the rice genomic sequence and
genetic information, and to identify and understand characteristics of genes, genome organization, pathways and phenotypes
in the cereals.
The exponential increase in the quantity and diversity of
information and the desire of the research community to query
and extract information from available data resources in a consistent manner demand standardized methods for describing,
classifying and inter-relating objects such as genetic markers
and loci, qualitative and quantitative phenotypes, polymorphisms, germplasms and sequences of both DNA and protein.
Users of the rice-specific databases including Gramene want to
be able to find and extract information in a way that allows
them to build useful biological statements based on comparative analyses. This demand from Gramene users has presented
two fundamental challenges in the management of diverse data
sets. One challenge involves the design and implementation of
an appropriate and robust database structure, and the second is
proper attribution of the data types, as a part of the annotation
process. The process of attribution is very important, given that
every data module in Gramene, such as the mutant, QTL and
protein, typically provides carefully collected information on
the phenotypes and deduced function of the gene product. As a
part of the attribution process, these modules link freely to
other databases, such as Oryzabase, GenBank (Benson et al.
2004) [see Appendix 1 (5)], Swiss-Prot (Boeckmann et al.
2003) [see Appendix 1 (6)], GenomeNet (Kanehisa 1997) [see
Appendix 1 (7)] and PubMed (Thirup and Nielsen 2002) [see
Appendix 1 (8)], to access supportive links for additional/original source information which was collected elsewhere. Appropriate attribution is critical in substantiating the validity and
original source of the data acquired by a third party, and it also
contributes an important dimension to the statements a user can
reliably make based on a database query.
Gramene, in collaboration with Oryzabase and various
other plant databases, such as MaizeGDB (Lawrence et al.
2004) [see Appendix 1 (9)], GrainGenes (Matthews et al. 2003)
[see Appendix 1 (10)], BarleyBase [see Appendix 1 (11)], IRIS
(Bruskiewich et al. 2003) [see Appendix 1 (12)] and TAIR
(Rhee et al. 2003) [see Appendix 1 (13)], has begun to work
together to develop common vocabularies and exchange protocols to enhance the capacity for data sharing, and to provide
plant researchers with the possibility of querying across various
plant databases. In the Gramene database, we have integrated
GO, PO and TO, i.e. gene, plant and trait ontologies.
The plant ontology (PO) in Gramene
The PO (Bruskiewich et al. 2002) provides a framework
for comparative collection of phenotypic information across
species by using a common vocabulary to describe morphologies and developmental stages of plants. It is also useful for
describing the tissue- or developmental stage-specific expression patterns of genes so far reported and gene profiles
obtained by microarray analyses. The development of a PO,
now in progress by members of the plant research community,
aims to provide a common, structured vocabulary for describing features of morphology, growth pattern and developmental
stages of flowering plants. Since the growth and developmental stage ontology is still under development at present by the
project in the community, Gramene curators have developed
the growth stage ontology (GRO) of cereals, in order to perform the data collection work appropriately as well as to provide a platform for collecting information and comparing
growth stages. The present GRO is in an initial stage such that
plant growth stages have been defined only for rice, maize, sorghum, oat, and plants in Triticeae such as wheat and barley. In
contrast, Oryzabase, as described below, has defined fine
developmental stages of rice in embryogenesis and vegetative
and reproductive phases and therefore this has been introduced
into the GRO [see Appendix 1 (14)]. We expect to provide
detailed information on the developmental stages of other
plants, which is associated with gene expression patterns and
mutant phenotypes, by the first quarter of 2005.
The TO was developed to provide a reliable framework
for describing the kinds of assays used to evaluate plant phenotypes, thus helping to standardize the descriptions of mutants,
strains, polymorphisms and QTLs. This is particularly relevant
since breeding and genetics communities that have a long history often use different terms for a certain trait or phenotype.
The TO developed by Gramene restructures and standardizes
methodology and terminology for evaluation of traits, which
are familiar to crop breeders around the world (SES: standard
evaluation system for rice) [see Appendix 1 (15)], GRIN [see
Appendix 1 (16)], ICIS [see Appendix 1 (17)] and IRIS, etc.).
Terms from the GO, PO and TO are all used to provide a coherent evaluation of a plant phenotype and function of a gene.
Their useful implementation in the Gramene database allows
the users to ask the following questions: find all mutant phenotypes evaluated for a trait ‘plant height’, and show which ones
have a sequenced gene associated with it. If there is a
sequenced gene, what is its (known/putative) molecular function and what biological process does it work in. In addition,
what is the location of the gene in the rice genome and are
there any known orthologs in other cereal genomes. It is easy
for the databases as well as the researchers to keep track of the
ontology terms associated with their gene or phenotype of
interest. This is accomplished by assigning a unique identifier
Gramene and Oryzabase
65
Fig. 1 Ontology map in Oryzabase. The six rectangles correspond to information categories which Oryzabase houses:
(i) genetic resources including mutants and wild relatives of
rice; (ii) gene dictionary/markers; (iii) linkage, physical and
comparative maps; (iv) catalogue of developmental stages
and gene expression patterns at those stages; (v) DNA
sequences; and (vi) references. Each shape (circle, diamond
and hexagon) shows the association of database contents with
each ontology, GO, TO and PO. By sharing ontology IDs,
Oryzabase has cross-references to Gramene and other external databases.
or an accession, similar to a GenBank accession number, to
each term in the ontology.
The gene ontology (GO) in Gramene
The GO is an example of a common vocabulary that has
been adopted by many genome databases to describe the
molecular function, role in a biological process and cellular
localization of the gene products from diverse organisms. As of
October 2004, Gramene provides information on molecular
characteristics of all the rice protein entries in SWISS-PROT
and TrEMBL. The information is extracted based on computational analyses as well as manual collection from peer-reviewed
publications. The annotations are shared with research communities through databases of the GO Consortium.
Ontologies and mutants in Gramene
The mutant database provides collected information on
mutant stocks of rice, which are publicly available. It includes
descriptions of phenotypes concerning morphological, developmental and agronomically important traits. The current version
of the mutant database houses information on >1,300 mutants,
of which about 400 mutants have information collected from
the literature and phenotype descriptions from Oryzabase and
from T. Kinoshita (personal communication).
Researchers can search the database using a gene symbol,
gene name or a Gramene accession number as query [see
Appendix 1 (18)]. Alternatively, the mutants can be searched
either by browsing gene symbols in an alphabetical order or by
searching the TO term by keywords to find the associated
mutant genes. For example, if you enter the word semidwarf-1
in the mutant search page of Gramene, you will find a list that
contains eight genes (loci) such as semidwarf-1 (sd1), semidwarf-11 and semidwarf-12. The list also includes gene symbols,
synonyms and brief descriptions of phenotypes. Then, if you
want to know details of semidwarf-1 (sd1), which has been
well characterized among the eight semi-dwarf genes, you can
move to the sd1 page that provides detailed information concerning this gene; name, allele, germplasm, more detailed
description of the phenotype together with two images, GenBank accession number, gene product, map position, associated features and literature references [see Appendix 1 (19)].
The associated feature section tells more about the phenotype’s description by using the concept of a controlled vocabulary (ontology). The TO term (culm length) indicates the trait
for which the gene was characterized, and the developmental
stage (04-stem elongation stage) and anatomy location (stem)
mean the stage and the part of the plant that the mutation
affects, respectively. The links from these controlled vocabulary terms take you to other ontology browsers, where you can
find the other mutant phenotypes that have been evaluated for
having the same trait, e.g. the trait ‘culm length’ [see Appendix 1 (20)]. Similarly, you can find other mutants that have
been known to express phenotype at a given developmental
stage or in a given plant part. The map position and sequence
information section provides information on alleles and their
genetic backgrounds in which the alleles were observed. The
map position section indicates a rough location of the gene on a
genetic map and this also links to the CMap pages. If the gene
corresponding to the mutation is cloned, you can obtain nucleotide and amino acid sequences and information on the gene
and protein through GenBank and Gene Product. In order to
learn more about the molecular characteristics of the protein
encoded by the gene, the link from Gene Product takes you to
detailed information.
Ontologies and QTLs in Gramene
In addition to the mutants, the Gramene database also provides information on the published QTL from rice, maize, oat,
barley and wild relatives of rice (Ni et al. in preparation) [see
Appendix 1 (21)]. The QTL information includes the follow-
66
Gramene and Oryzabase
Fig. 2 (a) Overview of GO structure produced by GOALL and hierarchical list of GO terms. A dot in the center indicates the root of GO term
hierarchy. The three organizing principles of GO are molecular function (blue–yellow), biological process (purple) and cellular component
(orange–red); each tree extends away from the center toward a different direction. All nodes are called ‘GO terms’ and each term has a unique
GO-ID. All nodes except the root have a parent node and some nodes have child node(s). Each GO-ID is unique, and more than two nodes can
share the same GO-ID because a term has relationships to one or more terms. For example, a gene product has one or more functions, is used in
one or more biological processes, and might be associated with one or more cellular components. (b) Comparison of the entire set of genes
between two species by GOALL. The details are described in the GOALL section of this mini-review.
ing. (i) Trait name, which is a controlled vocabulary term from
the TO. (ii) Trait symbol (e.g. SDCL, ALSN and DTHD)
which is assigned by Gramene in a controlled way. (iii) Trait
category based on agronomic concepts. A trait such as ‘Chalkiness of endosperm’ is classified into the category ‘quality’. The
categories are represented as top level parent terms in the TO.
Gramene and Oryzabase
The trait browser available from the QTL pages does not link
back to the TO, but instead displays only the number of QTL
associations with a trait. (iv) Information on the linkage group
and the position of the QTLs on the genetic map, as described
in the original articles.
An important organizational principle of the collected portion of all modules related to phenotype in Gramene is the utilization of ontologies to provide standardized terms for
describing phenotypes. This is critical, as it provides the backbone required to support the comparative querying functions
provided by the database.
Ontologies in Oryzabase
Oryzabase, which originated as a database of rice genetic
resources (established in 1995), has now grown into an integrated database of rice science. It provides a central location
for integration of rice genetic, genomic and phenotypic data,
and houses information such as (i) genetic resources including
mutants and wild relatives of rice; (ii) gene dictionary; (iii)
linkage, physical and comparative maps; (iv) catalogues of
developmental stages and gene expression patterns at those
stages; (v) DNA sequences; and (vi) basic information about
rice science for students. Among these contents, genes, mutant
phenotypes and stage-specific gene expression patterns are
directly related to the biological ontologies. The structure of
Oryzabase is schematically represented together with association with the biological ontologies (Fig. 1).
Oryzabase GO
There are three approaches concerning the GO in Oryzabase. The first one is the gene dictionary. The Oryzabase gene
dictionary is separated into two main groups, the genes that are
characterized biochemically and those related to phenotypic
traits. Only the former genes have assigned GO-IDs. The latter
genes, together with some genes in the former category, have
associations with the existing TO (refer to Kurata et. al. in this
issue). The second approach is to make GO associate predicted
genes on the rice genomic sequence. GO-IDs are assigned to
each predicted gene according to the sequence similarity by
BLAST analysis against the TrEMBL protein sequence database. The third approach is to develop a GO viewer, GOALL,
which enables users to search the vast amount of information in
the GO data set efficiently.
GOALL
As of October 2004, about 18,000 terms and >1,300,000
associated genes are available from the GO database. Fig. 2(a)
shows an overview of the structure of the GO world, i.e. a set
of GO-IDs or GO-terms. As shown in Fig. 2(a), all terms in the
GO as well as its hierarchy structure can be browsed with
GOALL. Users can also retrieve genes associated with a term
using the tool. GOALL provides both information from the GO
database and the Oryzabase gene dictionary.
67
GOALL has a characteristic function which allows comparison of the entire set of genes between two species. As
shown in Fig. 2(b), for example, a taxonomy data set is
uploaded and then the two species to be compared can be
selected from the pull-down menu, in this case Arabidopsis as
data A and rice as data B. The tool then searches all associated
genes in these two species and shows the terms which have at
least one associated gene. In this case, there are 1,486 GO
terms (green), indicating that associated genes are found in
both species, 1,694 terms (red) only found in Arabidopsi, 95
terms (blue) only found in rice and 14,436 terms not found in
either species. The number of terms reported here only reflects
the current status of gene annotation, not the actual biological
function. In order to support biologists to promote their studies
by GOALL, rapid progress in precise annotation of gene functions is needed [see Appendix 1 (22)].
Mutant collection in Oryzabase
The mutant collection of Oryzabase contains two classes
of mutant. One is morphologically and/or physiologically characterized mutants collected for a few decades. The other is a
collection from the large-scale mutant population of chemically mutagenized lines (Satoh and Omura 1981). The former
collection is classified into seven classes according to the gene
classification by Kurata et al. (this issue) and covers >2,000
mutants or variants. All these mutants are characterized with
several features including trait gene name, specific phenotype
and TO/PO/GO designations. The mutants in the latter collection are classified into three growth stages: seedling, vegetative and reproductive. In each stage, mutants are sorted further
into 50 categories according to the abnormalities in phenotype.
The Tos17 transposon-tagged lines generated by Hirochika et
al. (1996) in the Rice Genome Research Program also share the
latter mutant classification scheme. The number of these
mutants will grow to 10,000 lines in a few years.
Oryzabase
GENE EXPRESSION SPECIFIC FOR DEVELOPMENTAL STAGES AND
ORGANS
The definition of developmental stages and genes related
to development, which are described in detail in this issue (see
Itoh et al. and Kurata et al., this issue), are available through
Oryzabase. The defined stages in embryo, leaf, stomata, inflorescence, spikelet and ovule development are explained as
events characteristic of each stage together with images. The
list of enhancer trap lines, cell markers (genes expressed stage
specifically) and mutants, which covers all stages and all
organs, is also retrievable from the same web site. Although the
current release provides only a limited number of genes and
mutants, all items which appear in this issue will be incorporated into the database in the near future. A data submission
system for this content is being developed in order to encourage researchers to submit their data easily.
68
Gramene and Oryzabase
Future Direction of Biological Ontologies in Plants
Recent achievements in genomic science for Arabidopsis
have brought a new era into the biological ontology of plants.
According to the current release of GO, about 3,000 terms and
>80,000 associated genes are recorded in Arabidopsis (the estimated number of Arabidopsis genes is approximately 30,000;
associated genes in the GO database are derived from different
databases, resulting in redundancy) and there are almost half as
many in rice. With the completion of rice genome sequencing,
the number of annotated genes will be expected to increase rapidly in rice and these two model plants will become actual
anchors for dicot and monocot plant sciences.
Gramene will continue to work toward extending its annotation of rice gene, structure, trait and developmental stages in
collaboration with other plant databases. Ontology built in one
domain is desirable but it can be greatly improved by international collaborations. Accurate ontologies strongly rely on support from biological researchers. Biological researchers,
however, are not familiar with indices or conceptual hierarchy.
Also, biologists may prefer a one-stop database rather than
wandering through dispersed websites. To respond to these
demands, Oryzabase is now developing a new platform for
ontology, which is designed to display as many existing ontologies and ontology-related contents as possible without contradiction.
On the other hand, several new and challenging
approaches are underway in the bioinformatics community on
the development of automatic annotation techniques such as
text mining and information extraction. Ontologies facilitate
not only communication between researchers from different
fields but also the use of knowledge compiled by computers.
We are on the way to a future where new concepts and knowledge can be extracted from a vast amount of data by applying
ontologies.
Appendix 1
(1)
(2)
(3)
(4)
(5)
(6)
(7)
(8)
(9)
GO http://www.geneontology.org/
FlyBase http://flybase.bio.indiana.edu/
Gramene http://www.gramene.org/
Oryzabase http://www.shigen.nig.ac.jp/rice/oryzabase/
GenBank http://www.ncbi.nlm.nih.gov/
Swiss-Prot http://kr.expasy.org/sprot/
GenomeNet http://www.genome.jp/
PubMed http://www.ncbi.nlm.nih.gov/entrez/query.fcgi
MaizeGDB http://www.maizegdb.org/
(10)
(11)
(12)
(13)
(14)
(15)
(16)
(17)
(18)
(19)
(20)
(21)
(22)
GrainGenes http://wheat.pw.usda.gov/GG2/
BarleyBase http://www.barleybase.org/
IRIS http://www.iris.irri.org/
TAIR http://www.arabidopsis.org/
GRO http://www.gramene.org/db/ontology/search_term?id=GRO:
0007136
SES http://www.riceweb.org/ses/sesidx.htm
GRIN http://www.ars.grin.gov/npgs
ICIS: http://www.iris.irri.org/icis/SearchIRIS.htm
http://www.gramene.org/rice_mutant/Mutant_Search1.html
http://www.gramene.org/db/mutant/search_mutant?id=GR:
0060842
http://www.gramene.org/db/ontology/search_term?id=TO:
0000309
http://www.gramene.org/db/qtl/qtl_display
GOALL http://www.shigen.nig.ac.jp/ontology/
Acknowledgments
We thank Dr. David Mathews of Cornell University for critical
reading of the manuscript. Reviewers of the paper provided helpful
advice.
References
Benson, D.A., Karsch-Mizrachi, I., Lipman, D.J., Ostell, J. and Wheeler, D.L.
(2004) Nucleic Acids Res. 32: D23–D26.
Boeckmann, B., Bairoch, A., Apweiler, R., Blatter, M.C., Estreicher, A. et al.
(2003) Nucleic Acids Res. 31: D365–D370.
Bruskiewich, R., Coe, E.H., Jaiswal, P., McCouch, S., Polacco, M., Stein, L.,
Vincent, L. and Ware, D. (2002) Comp. Funct. Genomics 3: 137–142.
Bruskiewich, R.M., Cosico, A.B., Eusebio, W., Portugal, A.M., Ramos, L.M. et
al. (2003) Bioinformatics 19: i63–i65.
FlyBase Consortium (2003) Nucleic Acids Res. 31: 172–175.
Gene Ontology Consortium (2004) Nucleic Acids Res. 32: D258–D261.
Hirochika, H., Sugimoto, K., Otsuki, Y., Tsugawa, H. and Kanda, M. (1996)
Proc. Natl Acad. Sci. USA 93: 7783–7788.
Itoh, J.-I., Nonomura, K.-I., Ikeda, K., Yamaki, S., Inukai, Y., Yamagishi, H.,
Kitano, H. and Nagano, Y. (2005) Plant Cell Physiol. 46 (in this pci501).
Jaiswal, P., Ware, D., Ni, J., Chang, K., Zhao, W. et al. (2002) Comp. Funct.
Genomics 3: 132–136.
Kanehisa, M. (1997) Trends Biochem. Sci. 22: 442–444.
Kurata, N., Miyoshi, K., Nonomura, K.-I., Yamazaki, Y. and Ito., Y. (2005)
Plant Cell Physiol. 46 (in this pci506).
Lawrence, C.J., Dong, Q., Polacco, M.L., Seigfried, T.E. and Brendel, V. (2004)
Nucleic Acids Res. 32: D393–D397.
Matthews, D.E., Carollo, V.L., Lazo, G.R. and Anderson, O.D. (2003) Nucleic
Acids Res. 31: D183–D186.
Price, C.A. and Reardon, E.M. (2001) Nucleic Acids Res. 29: D118–D119.
Rhee, S.Y., Beavis, W., Berardini, T.Z., Chen, G., Dixon, D. et al. (2003)
Nucleic Acids Res. 31: D224–D228.
Satoh, H. and Omura, T. (1981) Jpn. J. Breed. 31: 316–326.
Thirup, P. and Nielsen, K.H. (2002) Ugeskr Laeger 164: 4553–4554.
Ware, D.H., Jaiswal, P., Ni, J., Yap, I.V. and Pan, X. (2002) Plant Physiol. 130:
1606–1613.
Yamazaki, Y., Nagato, Y. and Kurata, N. (2003) Rice Genet. Newslett. 20: 9–10.
(Received November 1, 2004; Accepted November 13, 2004)