Bioinformatics Advance Access originally published online on April 7, 2005
Bioinformatics 2005 21(19):3787-3793; doi:10.1093/bioinformatics/bti430
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
Automated genome annotation and pathway identification using the KEGG Orthology (KO) as a controlled vocabulary
1Center for Bioinformatics, National Laboratory of Protein Engineering and Plant Genetic Engineering, College of Life Sciences, Peking University Beijing 100871, P.R. China
2Biomedical Informatics, Department of Medicine, Stanford University School of Medicine Stanford, CA 94305, USA
*To whom correspondence should be addressed.
 |
Abstract |
|---|
 TOP Abstract INTRODUCTIONMETHODSRESULTSDISCUSSIONCONCLUSIONREFERENCES |
|---|
Motivation: High-throughput technologies such as DNA sequencing and microarrays have created the need for automated annotation of large sets of genes, including whole genomes, and automated identification of pathways. Ontologies, such as the popular Gene Ontology (GO), provide a common controlled vocabulary for these types of automated analysis. Yet, while GO offers tremendous value, it also has certain limitations such as the lack of direct association with pathways.
Results: We demonstrated the use of the KEGG Orthology (KO), part of the KEGG suite of resources, as an alternative controlled vocabulary for automated annotation and pathway identification. We developed a KO-Based Annotation System (KOBAS) that can automatically annotate a set of sequences with KO terms and identify both the most frequent and the statistically significantly enriched pathways. Results from both whole genome and microarray gene cluster annotations with KOBAS are comparable and complementary to known annotations. KOBAS is a freely available standalone Python program that can contribute significantly to genome annotation and microarray analysis.
Availability: Supplementary data and the KOBAS system are available at http://genome.cbi.pku.edu.cn/download.html
Contact: weilp{at}mail.cbi.pku.edu.cn
|
INTRODUCTION |
|---|
|
TOPAbstractINTRODUCTIONMETHODSRESULTSDISCUSSIONCONCLUSIONREFERENCES |
|---|
In recent years, high-throughput technologies such as DNA sequencing and microarrays have created the need for the automated annotation and analyses of large sets of genes, including whole genomes. To this end, an ontology, which is defined as a specification of a conceptualization, provides a common controlled vocabulary to facilitate electronic communication and sharing of information across different research groups and enables comparison of annotations across different genomes and different gene sets.
Several ontologies have been developed for genome annotation and expression analysis such as the Gene Ontology (GO) (Ashburner et al., 2000), GeneX (http://www.ncgr.org/genex/) and eVOC (Kelso et al., 2003). One of the most widely used ontologies is the Gene Ontology (GO). The Gene Ontology organizes functional terms into three top-level categories: molecular function, biological process and cellular component. Each category is structured as a directed acyclic graph (DAG) in which a term may have more than one parent and more than one child. The Gene Ontology has been used in the annotation of many genome databases, including SGD, CGD, FlyBase, MGI, TAIR, ZFIN, DictyBase, WormBase and RGD. Researchers annotating these databases use a combination of automation and manual curation to assign GO terms to genes in these genomes. Computational tools that have been developed include OntoBlast and Goblet, which assign GO terms to a new sequence based on its similarity (often measured by BLAST E-values) to a sequence with a known GO assignment (Hennig et al., 2003; Zehetner, 2003), and InterPro, which assigns GO terms to a new sequence based on the known GO assignments of the functional domains identified in the sequence (Mulder et al., 2003).
Other tools have been developed to discover significantly enriched GO terms among a given set of sequences such as a set of genes found to form a cluster in a microarray analysis. They include the web-based applications FatiGO (Al-Shahrour et al., 2004), GFinder (Masseroli et al., 2004), Gostat (Beissbarth and Speed, 2004), NetAffx GO Mining Tool (Cheng et al., 2004a), Onto-Express (Draghici et al., 2003), the JAVA application GoMiner (Zeeberg et al., 2003), the R package OntologyTraverser (Young et al., 2005) and the Perl command-line tools GeneMerge (Castillo-Davis and Hartl, 2003) and TermFinder (Boyle et al., 2004). These tools calculate the p-value for each GO term seen in the given set of sequences using one or more of the following statistical methods: binomial distribution, hypergeometric distribution, Fisher's exact test and/or chi-square test, using either the entire probe set on the microarray or the complete genome sequence for the background distribution. They also often apply one or more multiple hypotheses correction strategies, such as Bonferroni correction, to control the false discovery rate (FDR). Two other methods have used a knowledge-based approach (Cheng et al., 2004b) or DAG (directed acyclic graph) structure (Lee et al., 2004) to find meaningful GO terms.
While GO offers tremendous value, it also has certain limitations. Firstly, the GO hierarchy has highly varied depths along different branches—from two levels (e.g. GO:0001662 behavioral fear response) to 15 levels (e.g. GO:0030607 mitotic spindle orientation). Some of the variation is inherent in different functional families, while some may be an artifact of the uneven contribution by different groups participating in GO's development and may affect the reliability of statistical significance tests of GO terms. Secondly, because GO was originally developed for the annotation of eukaryotic genomes, the functional categorization in GO, and genome annotation using GO, is not as accurate for some prokaryotes as for eukaryotes. Thirdly, because GO terms do not correspond directly to known pathways; it is difficult to identify pathways directly from GO annotations.
We have investigated alternative controlled vocabularies for automated annotation of sets of genes and propose the KEGG Orthology (KO), part of the KEGG suite of resources (Kanehisa, 1997; Kanehisa and Goto, 2000) as such an alternative. KEGG is best known for its large pathway database and KO was developed to integrate pathway and genomic information in KEGG. Historically, enzyme commission (EC) numbers were used to describe common gene products in metabolic pathways. The ortholog identifiers were later introduced to overcome limitations in the enzyme nomenclature. The KEGG Orthology is a further extension of the ortholog identifiers, and is structured as a DAG hierarchy of four flat levels. The top level consists of the following five categories: metabolism, genetic information processing, environmental information processing, cellular processes and human diseases. The second level divides the five functional categories into finer sub-categories. The third level corresponds directly to the KEGG pathways, and the fourth level consists of the leaf nodes, which are the functional terms.
Our research demonstrates that KO is effective as a controlled vocabulary for automated annotation of sets of sequences, including whole genomes, and since KO links directly to known pathways, KO annotations enable concurrent pathway identification. Surprisingly, there are few published investigations using KO for automated annotations of new sequences, and the only existing tool is GFIT (Bono et al., 1998), which assigns EC numbers to query sequences based on orthologous genes in KEGG GENES. Unfortunately, the available GFIT package is obsolete and KO is significantly more complicated than EC numbers.
Several existing tools use the KEGG pathway database to identify enriched pathways in microarray data. PathProcessor (Grosu et al., 2002) and PathMAPA (Pan et al., 2003) map genes to KEGG pathways through their enzyme nomenclature, while ArrayXPath (Chung et al., 2004) maps to pathways in KEGG, GenMAPP (Dahlquist et al., 2002) and BioCarta (http://www.biocarta.com/), through sequence identifiers. Knowing the EC number or sequence identifier a priori is required to use these tools.
Here we present KOBAS, a KO-Based Annotation System written in Python that uses KO as a controlled vocabulary to automatically annotate, based on BLAST similarity searches, sets of new sequences. KOBAS also identifies the most frequent and the most significantly enriched pathways in a given set of sequences. We demonstrate the effectiveness of KO and KOBAS by evaluating KOBAS on two previously annotated genomes, one newly sequenced genome and gene clusters from a microarray experiment.
|
METHODS |
|---|
|
TOPAbstractINTRODUCTIONMETHODSRESULTSDISCUSSIONCONCLUSIONREFERENCES |
|---|
Selection and parsing of original datasets
We used two datasets available in Release 32.0 of the KEGG suite of resources. First, we extract the KO hierarchy and the known associations between genes and their corresponding KO functional terms from the KO data set. The KO dataset is a single complex flat file containing entries for all of the KO functional terms (the leaf nodes at the fourth level of the KO hierarchy). An example is shown in Supplementary Table 1. Each entry includes the ID, functional description and hierarchy location of the KO term, links to other databases (e.g. GO, COG, enzyme) and the IDs of genes in all annotated genomes that have this KO function. Second, we extract the amino acid sequences of all the annotated genes from the KEGG GENES dataset, which consists of one flat file for each annotated genome of gene entries. Each entry contains the gene's ID, definition, genomic location, KO assignments (if available), and amino acid and nucleotide sequence. An example is shown in Supplementary Table 2. About one-third of all sequences in the genomes in KEGG GENES have been assigned KO terms. We observed that while it was possible to extract known associations between genes and KO terms from both the KO and the KEGG GENES datasets, our detailed comparison of the set of associations extracted from the KO dataset versus that from the KEGG GENES dataset showed that they are not identical (Supplementary Table 3). In Release 32.0, the associations extracted from the KO dataset are an almost perfect superset of the associations extracted from the KEGG GENES dataset. Thus we chose to use the KO data set for the KO hierarchy and gene-KO associations and the KEGG GENES dataset for the amino acid sequences only.
We implemented a complex regular expression to parse the KO dataset and a series of regular expressions to parse the KEGG GENES dataset. Parsed information was converted into an intermediate XML format using the Simple API for XML (SAX) and then stored in an embedded relational database using SQLite.
From Release 32.0 we extracted 5540 KO leaf nodes of functional terms, 179 pathways, 748 177 genes (from 19 eukaryotes, 168 bacteria and 19 archaea), and 232 637 documented associations between genes and KO terms.
Automated annotation of new sequences with KO
Given a new DNA or protein sequence, we assign it KO terms based on its similarity to amino acid sequences in the KEGG GENES dataset. We calculate similarity using BLASTX for a new DNA sequence and BLASTP for a new protein sequence (Altschul et al., 1990, 1997). As choosing the right threshold is critical for achieving the most accurate assignments for the most sequences, we tested a variety of different thresholds and combinations of thresholds. In the end we chose as default,
 |
5 reduces the possibility of wrong KO assignments and, while it may miss some KO assignments, our evaluation and comparison results (presented in the Results section) indicate that the increase in precision exceeds the loss in sensitivity. These default thresholds can be easily customized by users of KOBAS. Any set of sequences, including whole genomes, can be annotated with the above procedure.
Identification of frequent and enriched pathways
It is often important to identify pathways involved in a set of sequences, such as a cluster of genes from microarray analysis. Since the third level in the KO hierarchy corresponds to KEGG pathways, once we find the right KO terms for a gene, we can trace back through the KO hierarchy to its associated pathways. Given a set of gene (or protein) sequences, we can identify the most frequently occurring pathways. For each pathway that occurs in the set of genes, we count the total number of genes in the set that are involved in the pathway and then rank the pathways by the number of genes.
Since some pathways are larger (i.e. involve more genes) than others, they tend to show up more frequently in any set of genes. Thus it is often valuable to identify the statistically significantly enriched pathways in a set of sequences, given a background distribution. We use the whole genome as the default background distribution (though users can define their own background distribution in the KOBAS system). For each pathway that occurs in the set of genes, we count the total number of genes in the set that are involved in the pathway. We then calculate the p value using a hypergeometric distribution. If a whole genome has N total genes, among which M are involved in the pathway under investigation, and the set of genes has n total genes, among which m are involved in the same pathway, the p value for the pathway is calculated as follows:
 | (1) |
KOBAS
We implemented the above methods in KOBAS. It is a standalone command-line program written in Python (2.3.4). It consists of three modules: kparser, blast2ko and pathfind. Kparser uses BioPython (1.3.0) and Martel (0.9.0) to parse the KO and KEGG GENES datasets. The parsed information is managed with SQLite, a small C library that implements a self-contained, embeddable and zero-configuration SQL database engine. Blast2ko automatically annotates a set of new sequences (in FASTA format) with KO functional terms. Pathfind identifies both the frequent and the enriched pathways in a given set of sequences. It calculates the FDR value by invoking the GeneTS (2.3) (Wichert et al., 2004) package of the R (2.00) language (Storey et al., 2002) through RPy, an interface from Python to R. KOBAS will run on most Linux systems, and executables are freely available at http://genome.cbi.pku.edu.cn/download.html.
|
RESULTS |
|---|
|
TOPAbstractINTRODUCTIONMETHODSRESULTSDISCUSSIONCONCLUSIONREFERENCES |
|---|
To evaluate the automated annotation of genes using KO as a controlled vocabulary, we first applied KOBAS to two whole genomes, one eukaryotic and one prokaryotic, that have existing KO annotations in KEGG, and compared our automated KO annotations with the existing KO annotations. We then applied KOBAS to a newly sequenced whole genome that has not yet been annotated in KEGG and compared our automated KO annotations with the textual annotations available in the NCBI genome database. To validate the use of KO annotations as an intermediate to identifying metabolic pathways, we applied KOBAS to 14 clusters of genes generated from a microarray experiment and compared our automatically identified pathways with the experimenters' original manual annotations. We also compared annotations of a well-annotated dataset using KO versus GO.
Evaluation of automated KO annotation of the Saccharomyces cerevisiae and Synechocystis genomes
To evaluate the automated annotation of genes using KO as a controlled vocabulary, we applied KOBAS to annotate the complete sets of genes in a eukaryotic genome, S.cerevisiae, and a prokaryotic genome, Synechocystis sp. PCC6803. Both genomes are well annotated in KEGG, although not all the genes have been annotated with KO terms, with 1478 (24%) of 6183 total genes in the S.cerevisiae genome and 1202 (36%) of 3314 total genes in the Synechocystis genome having existing KO annotations. To ensure a rigorous evaluation, we always remove the genome under study from the original KEGG datasets before applying KOBAS.
We used different evaluation criteria for genes with and without existing KO annotations. For genes with existing KO annotations in KEGG, we used their existing KO annotations as the gold standard for comparison with the automated annotations by KOBAS. We used two measures of annotation quality, precision and coverage, defined as follows:
 | (2) |
 | (3) |
There is a tradeoff between high precision and high coverage. We tested a variety of thresholds, including identity, E-value and rank, and found that a combination of E-value and rank cutoffs gives the best overall result. Figure 1 shows the precision and coverage of KOBAS' annotation of the S.cerevisiae and Synechocystis genomes using different E-value and rank cutoffs. With the default threshold, E-value < 10–5 and rank 5, the precision and coverage are respectively 94.3 and 90.3% for S.cerevisiae and 96.2 and 91.8% for Synechocystis.
|
For genes that have no existing KO annotations in KEGG, we manually compared KOBAS' annotations to the functional descriptions in other databases such as SGD (the Saccharomyces genome database, http://www.yeastgenome.org/). We found that KOBAS is capable of automatically creating meaningful annotations not available in KEGG. Several examples are given in Table 1.
|
Automated KO annotation of the new Propionibacterium acnes genome
We then applied KOBAS to a newly sequenced whole genome that is not yet annotated in KEGG: P.acnes, a commensal microbe of the human skin (Bruggemann et al., 2004). KOBAS automatically annotated 1177 (51%) of a total of 2297 genes in the P.acnes genome. This percentage is high compared to the percentages of annotated genes versus all genes in the annotated whole genomes currently in KEGG. Figure 2 shows a sample of KOBAS' annotation of P.acnes. We manually compared KOBAS' annotations with the functional annotations in the NCBI genome database, and found most of KOBAS' annotations to be correct. In particular, KOBAS was able to automatically identify a number of genes known to be important in P.acnes' ability to cope with changing oxygen tension. Examples include oxidative phosphorylation (PPA175-176), F-type H+-transporting ATPases (PPA1238–1243 and PPA1245), nitrate reductase (PPA507–510), anaerobic dimethyl sulfoxide reductases (PPA515 and PPA517) and glycerol-3-phosphate dehydrogenase (PPA2248–2250). Using the pathway identification function, KOBAS automatically discovered the most frequent pathways in P.acnes, as shown in Table 2, including the critical oxidative phosphorylation and nitrogen metabolism pathways. These findings are consistent with what is known about the biology of P.acnes (Bruggemann et al., 2004).
|
|
Identification of frequent and enriched pathways in microarray gene clusters
To evaluate KOBAS' ability to automatically identify frequent and enriched pathways in a set of genes, we applied it to the clusters of genes resulting from a recently published microarray experiment (Saldanha et al., 2004). In their experiment, Saldanha et al. studied the physiological response of S.cerevisiae to the limitation of different nutrients in batch and steady-state (chemostat) cultures using a whole-genome microarray. They identified 14 clusters of co-expressed genes and manually annotated the biological relevance of each cluster. For each cluster, we applied KOBAS to first annotate all the genes with KO and to then identify both the most frequent and the statistically significantly enriched pathways. With the rather strict cutoff of FDR < 0.05, KOBAS found statistically significantly enriched pathways in four of the clusters, as shown in Table 3. These automated results are consistent with the authors' annotations of the clusters in the original paper (Saldanha et al., 2004).
|
Comparison of KO and GO annotations on a well-annotated prokaryotic protein set
We compared the KO versus GO annotations of the important photosystems I and II protein family in Anabaena (PCC 7120). The dataset consists of 46 proteins, extracted from CyanoBase (http://www.kazusa.or.jp/cyano/) and carefully manually curated. It is available for download at http://genome.cbi.pku.edu.cn/download.html. We annotated the proteins with KO and then with GO. For KO annotations, we used KOBAS with the default parameters, as described in the Methods section. For GO annotations, we used one of the most popular GO annotation approaches which is to BLAST the query sequence against GOA UniProt sequences and select the GO terms of the top hit. KO annotated 44 out of the 46 proteins and all annotations were correct, whereas GO annotated 33 proteins with 28 correct.
|
DISCUSSION |
|---|
|
TOPAbstractINTRODUCTIONMETHODSRESULTSDISCUSSIONCONCLUSIONREFERENCES |
|---|
KOBAS is the first open-access system to use KO as a controlled vocabulary to automatically annotate a set of sequences, such as the complete set of genes in a whole genome or clusters of genes resulting from microarray analysis, and to identify both frequent and significantly enriched pathways. Evaluation results from both whole genomes and microarray gene clusters demonstrate the effectiveness of KOBAS. KOBAS is a standalone application written in Python and is highly modular, making it easy to incorporate into other applications either in part or as a whole.
In transferring KO annotations from known genes to new genes, we used a combination of E-value and rank as a threshold rather than E-value alone. Because only one-third to half of the genes in whole genomes currently have existing KO (or GO) annotations, we intentionally generated the BLASTable dataset using the collection of all gene sequences in whole genomes in KEGG, rather than using just those with KO annotations. Thus, for example, should a new gene be found to be similar to 100 top BLAST hits without KO annotations and a 101st hit with KO annotation, we would not assign this KO annotation to the new gene. It is more likely that the new gene is homologous to one of the top 100 genes that have not yet been annotated, rather than the 101st gene that happens to have been annotated. We found that by doing this, even though the total number of KO assignments made by KOBAS is lower, the number of false assignments is also dramatically lower, and we chose to value quality over quantity. Users can easily change this threshold in KOBAS to fit their own analysis goals. In future work we will consider adding more parameters such as percentage of alignment length over the length of the whole genes and concurrence of functional domains, and test their effect on precision and coverage.
A key function of KOBAS is automated identification of pathways in a set of genes or proteins and linking the genes directly to KEGG pathways. KOBAS identifies both the most frequent and the significantly enriched pathways. In determining the latter, KOBAS can use the whole genome, the whole probe set or any user-defined set of genes as the background distribution, and we implemented FDR correction to reduce the number of false positives introduced by multiple hypothesis testing. It is important to note that existing pathway identification systems, such as PathProcessor (Grosu et al., 2002), PathMAPA (Pan et al., 2003) and ArrayXPath (Chung et al., 2004) map genes to pathways based on EC numbers or gene identifiers and thus require that the genes already be in the KEGG pathway database. In contrast, KOBAS uses its automated KO annotation module to identify pathways in sets of genes that are not yet annotated in KEGG, making it especially valuable when analyzing genes in a newly sequenced genome or custom-made cDNA arrays.
Our results indicate that KO is effective as a controlled vocabulary and, although it would be premature to make general conclusions about the power of KO versus GO annotations without further large-scale, in-depth comparisons, some advantages of KO can be observed. At least for some biological systems, such as the important photosystems in the prokaryote genome we tested, KO seemed to provide better annotations in terms of both the total number of proteins annotated and the quality of the annotations. This may be due to the fact that historically KEGG has had a stronger focus on the prokaryotic species than has GO. The biggest limitation of KO is perhaps the fact that it does not yet have as many functional terms in its hierarchy as does GO, although this is improving as KO continues to grow. Overall, we believe that KO-based annotations are complementary to GO-based annotations and systems like KOBAS can make a significant contribution to genomic and proteomic analysis.
|
CONCLUSION |
|---|
|
TOPAbstractINTRODUCTIONMETHODSRESULTSDISCUSSIONCONCLUSIONREFERENCES |
|---|
We developed a KO-Based Annotation System (KOBAS) that can automatically annotate a set of sequences and identify both the most frequent and the statistically significantly enriched pathways. Results from both whole genome and microarray gene cluster annotations with KOBAS are comparable and complementary to existing annotations and demonstrate the use of KO as an alternative controlled vocabulary for automated annotation and pathway identification. KOBAS is a freely available standalone Python program that can contribute significantly to genome annotation and microarray analysis.
|
Acknowledgments |
|---|
This work was supported by the China National High-tech 863 Program (2004AA231020, 2004BA711A21). We thank Sheng Wei for helpful comments.
|
Footnotes |
|---|
The authors wish it to be known that, in their opinion, the first two authors should be regarded as joint First Authors. 
Received on January 18, 2005; revised on March 26, 2005; accepted on April 4, 2005
|
REFERENCES |
|---|
|
TOPAbstractINTRODUCTIONMETHODSRESULTSDISCUSSIONCONCLUSIONREFERENCES |
|---|
Al-Shahrour, F., et al. (2004) FatiGO: a web tool for finding significant associations of Gene Ontology terms with groups of genes. Bioinformatics, 20, 578–580
Altschul, S.F., et al. (1990) Basic local alignment search tool. J. Mol. Biol., 215, 403–410[CrossRef][Web of Science][Medline].
Altschul, S.F., et al. (1997) Gapped BLAST and PSI-BLAST: a new generation of protein database search programs. Nucleic Acids Res., 25, 3389–3402
Ashburner, M., et al. (2000) Gene ontology: tool for the unification of biology. The Gene Ontology Consortium. Nat. Genet., 25, 25–29[CrossRef][Web of Science][Medline].
Beissbarth, T. and Speed, T.P. (2004) GOstat: find statistically overrepresented Gene Ontologies within a group of genes. Bioinformatics, 20, 1464–1465
Benjamini, Y. and Hochberg, Y. (1995) Controlling the false discovery rate: a practical and powerful approach to multiple testing. J. R. Statist. Soc. B, 57, 289–300.
Bono, H., et al. (1998) Systematic prediction of orthologous units of genes in the complete genomes. Genome Inform Ser Workshop Genome Inform., 9, 32–40[Medline].
Boyle, E.I., et al. (2004) GO::TermFinder—open source software for accessing Gene Ontology information and finding significantly enriched Gene Ontology terms associated with a list of genes. Bioinformatics, 20, 3710–3715
Bruggemann, H., et al. (2004) The complete genome sequence of Propionibacterium acnes, a commensal of human skin. Science, 305, 671–673
Castillo-Davis, C.I. and Hartl, D.L. (2003) GeneMerge—post-genomic analysis, data mining, and hypothesis testing. Bioinformatics, 19, 891–892
Cheng, J., et al. (2004a) NetAffx gene ontology mining tool: a visual approach for microarray data analysis. Bioinformatics, 20, 1462–1463
Cheng, J., et al. (2004b) A knowledge-based clustering algorithm driven by Gene Ontology. J. Biopharm. Statist., 14, 687–700[CrossRef][Medline].
Chung, H.J., et al. (2004) ArrayXPath: mapping and visualizing microarray gene-expression data with integrated biological pathway resources using scalable vector graphics. Nucleic Acids Res., 32, W460–W464
Dahlquist, K.D., et al. (2002) GenMAPP, a new tool for viewing and analyzing microarray data on biological pathways. Nat. Genet., 31, 19–20[CrossRef][Web of Science][Medline].
Draghici, S., et al. (2003) Onto-Tools, the toolkit of the modern biologist: Onto-Express, Onto-Compare, Onto-Design and Onto-Translate. Nucleic Acids Res., 31, 3775–3781
Grosu, P., et al. (2002) Pathway processor: a tool for integrating whole-genome expression results into metabolic networks. Genome Res., 12, 1121–1126
Hennig, S., et al. (2003) Automated gene ontology annotation for anonymous sequence data. Nucleic Acids Res., 31, 3712–3715
Kanehisa, M. (1997) A database for post-genome analysis. Trends Genet., 13, 375–376[CrossRef][Web of Science][Medline].
Kanehisa, M. and Goto, S. (2000) KEGG: kyoto encyclopedia of genes and genomes. Nucleic Acids Res., 28, 27–30
Kelso, J., et al. (2003) eVOC: a controlled vocabulary for unifying gene expression data. Genome Res., 13, 1222–1230
Lee, S.G., et al. (2004) A graph-theoretic modeling on GO space for biological interpretation of gene clusters. Bioinformatics, 20, 381–388
Masseroli, M., et al. (2004) GFINDer: Genome Function INtegrated Discoverer through dynamic annotation, statistical analysis, and mining. Nucleic Acids Res., 32, W293–W300
Mulder, N.J., et al. (2003) The InterPro Database, 2003 brings increased coverage and new features. Nucleic Acids Res., 31, 315–318
Pan, D., et al. (2003) PathMAPA: a tool for displaying gene expression and performing statistical tests on metabolic pathways at multiple levels for Arabidopsis. BMC Bioinform., 4, 56[CrossRef][Medline].
Saldanha, A.J., et al. (2004) Nutritional homeostasis in batch and steady-state culture of yeast. Mol. Biol. Cell., 15, 4089–4104
Storey, J.D. (2002) A direct approach to false discovery rates. J.R. Statist. Soc. B, 64, 479–498[CrossRef].
Wichert, S., et al. (2004) Identifying periodically expressed transcripts in microarray time series data. Bioinformatics, 20, 5–20
Young, A., et al. (2005) OntologyTraverser: an R package for GO analysis. Bioinformatics, 21, 275–276
Zeeberg, B.R., et al. (2003) GoMiner: a resource for biological interpretation of genomic and proteomic data. Genome Biol., 4, R28[CrossRef][Medline].
Zehetner, G. (2003) OntoBlast function: from sequence similarities directly to potential functional annotations by ontology terms. Nucleic Acids Res., 31, 3799–3803
CiteULike 
Connotea 
Del.icio.us What's this?
This article has been cited by other articles:
|
|
 |
|
  C. Li, X. Li, Y. Miao, Q. Wang, W. Jiang, C. Xu, J. Li, J. Han, F. Zhang, B. Gong, et al. SubpathwayMiner: a software package for flexible identification of pathways Nucleic Acids Res., October 1, 2009; 37(19): e131 - e131. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 L. Yang, L. Xu, and L. He A CitationRank algorithm inheriting Google technology designed to highlight genes responsible for serious adverse drug reaction Bioinformatics, September 1, 2009; 25(17): 2244 - 2250. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 L. Wissler, E. Dattolo, A. D. Moore, T. B. H. Reusch, J. L. Olsen, M. Migliaccio, E. Bornberg-Bauer, and G. Procaccini Dr. Zompo: an online data repository for Zostera marina and Posidonia oceanica ESTs Database, August 4, 2009; 2009(0): bap009 - bap009. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 G. Wei, Y. Tao, G. Liu, C. Chen, R. Luo, H. Xia, Q. Gan, H. Zeng, Z. Lu, Y. Han, et al. Inaugural Article: A transcriptomic analysis of superhybrid rice LYP9 and its parents PNAS, May 12, 2009; 106(19): 7695 - 7701. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
C.-Y. Li, Q.-R. Liu, P.-W. Zhang, X.-M. Li, L. Wei, and G. R. Uhl OKCAM: an ontology-based, human-centered knowledgebase for cell adhesion molecules Nucleic Acids Res., January 1, 2009; 37(suppl_1): D251 - D260. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
J.-T. Li, Y. Zhang, L. Kong, Q.-R. Liu, and L. Wei Trans-natural antisense transcripts including noncoding RNAs in 10 species: implications for expression regulation Nucleic Acids Res., September 1, 2008; 36(15): 4833 - 4844. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 P. May, S. Wienkoop, S. Kempa, B. Usadel, N. Christian, J. Rupprecht, J. Weiss, L. Recuenco-Munoz, O. Ebenhoh, W. Weckwerth, et al. Metabolomics- and Proteomics-Assisted Genome Annotation and Analysis of the Draft Metabolic Network of Chlamydomonas reinhardtii Genetics, May 1, 2008; 179(1): 157 - 166. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 N. Novershtern, Z. Itzhaki, O. Manor, N. Friedman, and N. Kaminski A Functional and Regulatory Map of Asthma Am. J. Respir. Cell Mol. Biol., March 1, 2008; 38(3): 324 - 336. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 G. Sha, D. Wu, L. Zhang, X. Chen, M. Lei, H. Sun, S. Lin, and J. Lang Differentially expressed genes in human endometrial endothelial cells derived from eutopic endometrium of patients with endometriosis compared with those from patients without endometriosis Hum. Reprod., December 1, 2007; 22(12): 3159 - 3169. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
T. T. Tran, P. Dam, Z. Su, F. L. Poole II, M. W. W. Adams, G. T. Zhou, and Y. Xu Operon prediction in Pyrococcus furiosus Nucleic Acids Res., January 12, 2007; 35(1): 11 - 20. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
W. Zhang, Y. Zhang, H. Zheng, C. Zhang, W. Xiong, J. G. Olyarchuk, M. Walker, W. Xu, M. Zhao, S. Zhao, et al. SynDB: a Synapse protein DataBase based on synapse ontology Nucleic Acids Res., January 12, 2007; 35(suppl_1): D737 - D741. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
S. Li, L. Wu, and Z. Zhang Constructing biological networks through combined literature mining and microarray analysis: a LMMA approach Bioinformatics, September 1, 2006; 22(17): 2143 - 2150. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
Y. Zhang, X. S. Liu, Q.-R. Liu, and L. Wei Genome-wide in silico identification and analysis of cis natural antisense transcripts (cis-NATs) in ten species Nucleic Acids Res., July 18, 2006; 34(12): 3465 - 3475. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
J. Wu, X. Mao, T. Cai, J. Luo, and L. Wei KOBAS server: a web-based platform for automated annotation and pathway identification. Nucleic Acids Res., July 1, 2006; 34(Web Server issue): W720 - W724. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 Y.-H. Shi, S.-W. Zhu, X.-Z. Mao, J.-X. Feng, Y.-M. Qin, L. Zhang, J. Cheng, L.-P. Wei, Z.-Y. Wang, and Y.-X. Zhu Transcriptome Profiling, Molecular Biological, and Physiological Studies Reveal a Major Role for Ethylene in Cotton Fiber Cell Elongation PLANT CELL, March 1, 2006; 18(3): 651 - 664. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||