Skip Navigation
Pathway Interaction Database homepage

Bioinformatics Primer

Understanding environment-disease connections: An introduction to the Comparative Toxicogenomics Database (CTD)

doi:10.1038/pid.2011.2

The Comparative Toxicogenomics Database (CTD; http://ctd.mdibl.org) is a freely available database that is dedicated to promoting the exploration and development of testable hypotheses about the effects of the environment on human health. CTD integrates manually curated, cross-species information about gene environment (G×E) interactions and environment disease and gene disease relationships with public datasets such as functional annotations, molecular pathways and literature. Analytical and statistical tools are provided to facilitate access to and interpretation of CTD data. CTD is a unique resource for enhancing understanding about the mechanisms of chemical actions and the etiologies of environmentally influenced diseases.

The etiology of most chronic diseases involves interactions between the environment and genes that regulate important physiological processes. Many complex diseases are caused by reversible behaviors or avoidable exposures, whereas relatively few diseases are attributed only to genetic mutations. Environmental factors are implicated in many common conditions such as asthma, cancer and diabetes; however, the extent to which these factors and an individual's genetic composition contribute to the etiology of these widespread diseases remains unclear. To enhance understanding about the effects of the environment on human health, we developed the freely available CTD.1, 3 CTD provides manually curated data from the peer-reviewed scientific literature describing (i) cross-species chemical gene interactions from vertebrates and invertebrates, (ii) chemical disease relationships and (iii) gene disease relationships (Fig. 1). For the purpose of this report, we refer to genes and their products (e.g., proteins, messenger RNA) in the context of CTD as simply genes. By integrating these data with external datasets such as Gene Ontology (GO) annotations and pathway data from KEGG and Reactome resources, CTD enables development of novel, testable hypotheses about relationships among chemicals, genes, diseases, biological processes and molecular pathways.4, 5, 6 CTD is unique among other publicly available databases by virtue of its environmental focus (although therapeutic drugs are also included), manually curated datasets, integration of external datasets, and novel analytical tools.

Figure 1
Figure 1 :

CTD data content. CTD integrates curated data for chemical-gene interactions and chemical- and gene-disease relationships (colored circles) with select public datasets (gray circles; pathways from the KEGG and Reactome databases and GO annotations) to provide novel insights into the complex connections between the environment and human health. Solid lines describe directly curated or integrated relationships and dashed lines describe inferred relationships.


CTD data curation is prioritized largely by chemicals of interest as defined by entities such as the US Environmental Protection Agency (http://www.epa.gov/), the National Institute of Environmental Health Sciences (http://www.niehs.nih.gov/) and the US Food and Drug Administration (http://www.fda.gov/), as well as by collaborating scientists. Currently, CTD contains over 300,000 curated interactions involving over 6,000 chemicals and 19,000 genes in 330 organisms. There are over 11,000 and 13,000 curated chemical disease and gene disease relationships, respectively. A complete summary of CTD data content is shown in Table 1 and is updated monthly online (http://ctd.mdibl.org/about/dataStatus.go). The purpose of this report is to illustrate how CTD data can provide insights into the modes of chemical actions and how the environment may contribute to human disease using a range of query and analysis options. The discussion will emphasize how data are presented from a chemical perspective using the insecticide, chlorpyrifos as an example.


Searching CTD

The major datasets in CTD include chemicals, genes and diseases (Fig. 1). These entities are represented using standardized nomenclatures, which enable their unique integration with other external datasets. CTD data can be queried from the perspective of a chemical, gene, disease, GO term or pathway using quick- or advanced-search options (Fig. 2). A quick-search field is provided in the upper right-hand corner of most pages. Several advanced-query forms that allow construction of complex queries are provided under the “Search” tab (Fig. 2). For example, the “Chemical Gene Interaction Query” (Fig. 3) allows users to combine many search terms to retrieve data such as “Show me all curated interactions for estrogens that affect development” (chemical: estrogens; GO term: developmental process) or “Show me all curated interactions involving transcriptional effects associated with arsenic” (chemical: arsenic; chemical gene interaction: expression). All data in CTD are deeply integrated, and users can toggle easily between different data contexts (e.g., chemical or gene). Context-specific CTD datasets can be saved in various formats (.CSV, Excel, .XML, .TSV) by clicking on the download links at the bottom of each page, whereas full datasets are available on the “Downloads” tab of the home page (http://ctd.mdibl.org/downloads/). Customized datasets can be downloaded using the Batch Query tool (see below).

Figure 2
Figure 2 :

CTD home page. CTD is a freely available web-based resource. The CTD home page describes the database content, highlights recently updated datasets and news, and provides a keyword search mechanism and access to advanced-query forms (red boxes) and portals to the data resources and available tools (arrows).


Figure 3
Figure 3 :

Chemical Gene Interaction Query form. Users may frame complex queries in search of chemical gene interactions by using multiple fields in this query form. For example, specifying estrogens as a chemical and developmental process in the Gene Ontology field will retrieve all interactions in which estrogen affects genes involved in development. Specifying arsenic as a chemical and selecting “expression” as a chemical gene interaction will retrieve chemical gene interactions involving transcriptional effects of arsenic.


Chemicals

CTD contains curated data for over 6,000 diverse chemicals from pharmaceuticals to environmental contaminants. Chemicals are curated, stored and presented using the “Chemicals and Drugs” branch of the Medical Subject Headings (MeSH) vocabulary (http://www.nlm.nih.gov/mesh/).7 This vocabulary is hierarchical, allowing users to query by very specific (e.g., cadmium) or more general chemical terms (e.g., heavy metals). Most data pages for chemicals (except ChemComps) include information about the chemical of interest or any of its child terms. For each chemical, associated data are provided and include: Basic information, Interactions, Genes, Diseases, ChemComps, Pathways, GO, References and Links. These data are accessible via the nine corresponding data tabs on a chemical page and are described in detail below.

Basic. This tab provides an overview of the chemical of interest including the name, synonyms, CAS registry number, definition, chemical structure, link to the corresponding chemical entry in MeSH, graph of the top 10 curated interacting genes, and the position of the chemical within the MeSH hierarchy (Fig. 4). Here, a user could easily determine that chlorpyrifos is “an organothiophosphate cholinesterase inhibitor that is used as an insecticide and as an acaricide,” and the most abundantly curated interacting genes indicate that this compound affects acetylcholinesterase (ACHE) and several metabolic pathways.

Figure 4
Figure 4 :

Basic chemical information page. Every chemical in CTD has a Basic information page that provides the following details (where available): chemical name, CAS type 1 name, synonyms (equivalent terms), CAS registry number, definition, chemical structure, MeSH ID and link to the corresponding MeSH page, top 10 interaction genes and the position of the chemical within the chemical vocabulary hierarchy.


Interactions. This tab provides associated curated interactions for a chemical and any of its child terms (Fig. 5). For example, interactions for the chlorpyrifos metabolite, O,O-diethyl O-3,5,6-trichloro-2-pyridyl phosphate (chlorpyrifos oxon), and the closely related insecticide, chlorpyrifos-methyl, are included on the chlorpyrifos Interactions page. Interactions are captured using an action ontology developed for CTD curation.2 Data presented include the name of the chemical, the interacting genes, the interactions and links to the associated reference(s) and organism(s) in which the interactions were reported. The noted chlorpyrifos compounds have 344 curated interactions from multiple species, involving 233 genes. The most common of the 13 interaction types involves effects on gene transcription. The other chlorpyrifos interactions involve protein activity, binding, chemical synthesis, export, hydrolysis, hydroxylation, metabolic processing, oxidation, phosphorylation, response to chemical, secretion and sulfation.

Figure 5
Figure 5 :

Chemical Interactions page. Curated chemical gene interactions are displayed for chemicals, genes and diseases. Shown here is an interactions page for the chemical, chlorpyrifos, which includes: the interacting chemical or its child terms (in this example, data for the metabolite, O,O-diethyl O-3,5,6-trichloro-2-pyridyl phosphate, and the related compound, chlorpyrifos-methyl, are also included), the interacting genes, the interactions and links to the source reference(s) and organism(s) in which the interactions were reported (red arrows).


Genes. This tab lists all of the genes that have manually curated interactions with the chemical (or any of its child terms) in the “Interacting Gene” column (Fig. 6). The numbers of curated interactions in distinct organisms are provided in the “Interactions” column. By default, genes are ranked by the number of curated interactions. In this example, ACHE has the greatest number of curated chlorpyrifos interactions. Notably, these interactions were observed in 9 different organisms ranging from crustaceans to mammals, suggesting that chlorpyrifos may be acting through an evolutionarily conserved mechanism.

Figure 6
Figure 6 :

Interacting Genes page. Curated chemical gene interactions are displayed for chemicals, genes and diseases. Shown here is an Interacting Genes page for the chemical, chlorpyrifos, which includes: the interacting genes ranked by the number of curated interactions (red arrows).


Diseases. CTD contains curated and inferred chemical disease and gene disease relationships. Curated, or direct chemical–disease relationships are extracted from the published literature by CTD curators. Inferred relationships are established computationally via genes that have curated interactions with a chemical and a disease (i.e., chemical A is associated with disease C because chemical A has a curated interaction with gene B, and gene B has a curated relationship with disease C). The data presented on a chemical disease tab include the chemical of interest or its child terms, the disease associated with the chemical, the type of evidence for the relationship (if curated, M: Marker/mechanism and/or T: Therapeutic), the genes that form the basis of an inferred relationship, an inference score, and a link to the source reference(s) for the curated and inferred relationships (Fig. 7).

Figure 7
Figure 7 :

Chemical Disease page. Chemical disease relationships are directly curated or inferred by common interacting genes. Data on Chemical Disease pages include the chemical of interest or its child terms, the disease associated with the chemical, the evidence for the relationship (if curated, M: marker/mechanism and/or T: therapeutic), the genes that form the basis of an inferred relationship, an inference score, and a link to the source reference(s) for the curated and inferred relationships. Neurological diseases and cancers predominate among curated and inferred diseases associated with chlorpyrifos (red rectangles). Inference Networks include the set of genes that have curated interactions with both the chemical and the disease. This gene set is unique to CTD and provides insights into potential molecular mechanisms underlying the chemical disease relationship.


The inference score reflects the degree of similarity between CTD chemical–gene–disease networks and a similar scale-free random network. Many biological networks, such as disease and metabolic networks, have been shown to be scale-free random networks.9 The score takes into account the connectivity of the chemical, disease and each of the genes used to make the chemical disease inference. The higher the score, the more likely the inference network has nonuniform connectivity as observed in scale-free random networks.

A broad spectrum of diseases is associated with chlorpyrifos. Among others, curated diseases include respiratory tract diseases, neurotoxicity syndromes and agricultural workers' diseases, as might be predicted by its use as an insecticide with agricultural applications and by the curated interacting genes and GO enrichment results in CTD (Fig. 7). When sorting by inference score, the highest ranked inferred diseases partition largely into neurological disorders (autistic spectrum disorder, schizophrenia, Parkinson's disease, bipolar disorder) and cancers (prostate, breast and lung). The neurological disorders are consistent with chlorpyrifos' role as a cholinesterase inhibitor but also align with what is known about its comparable chemicals (e.g., clozapine is an antipsychotic treatment for schizophrenia and Parkinson's disease). The genes that form the basis of these inferred relationships may provide important insights into the mechanisms by which chlorpyrifos may modulate diseases in exposed individuals or populations or function as diagnostic biomarkers of exposure.

ChemComps. In relation to a chemical of interest, a ChemComps page provides the chemicals with the 20 most comparable sets of interacting genes, or by extension, the most similar potential modes of action. Similarity is measured by the Jaccard index, which is calculated as the size of the intersection of interacting genes for chemical A and chemical B divided by the size of the union of those genes.8 Overlapping gene sets are provided. Diazinon ranks as the most similar compound to chlorpyrifos in CTD, sharing 151 interacting genes (Fig. 8). Notably, among the chemicals with the top 10 similarity scores (excepting the less specific “plant extracts”): All are known neuromodulators; five are insecticides that function as cholinesterase inhibitors (diazinon, dieldrin, permethrin, parathion, DDT); one is a metal (nickel); and three are natural or synthetic compounds with dopaminergic effects (amphetamine, dopamine and clozapine). The overlap among interacting genes for these compounds can now be visualized by downloading and opening a “pathway view” file using Cytoscape (http://www.cytoscape.org/), an open-source pathway visualization application (Fig. 9). This network demonstrates that although there is substantial overall similarity between the chemicals, there appear to be distinct clusters (e.g., amphetamine, dopamine and clozapine) of interacting genes that may reflect unique aspects of their modes of action.

Figure 8
Figure 8 :

ChemComps page. ChemComps provide the 20 chemicals with the most comparable set of interacting genes, or by extension, the most similar potential modes of action, as the chemical subject of the page. Similarity (or comparability) is measured by the Jaccard index, which is calculated as the size of the intersection of interacting genes for chemical A and chemical B divided by the size of the union of those genes (red arrow). Similar chemicals and overlapping gene sets are provided (red arrows). Overlapping interacting genes for the top 10 chemicals may be evaluated in a pathway view by downloading and opening an XGMML-formatted file in Cytoscape (http://www.cytoscape.org/), an open-source pathway visualization software, or by opening the network view online (red rectangle).


Figure 9
Figure 9 :

ChemComps network view. Network views of comparable chemicals illustrate the overlap among the compounds' interacting genes. Evident here is significant overlap among the interacting genes for the 10 chemicals most like chlorpyrifos. The top three ranked compounds (Diazinon, dieldrin and nickel) and the dopaminergic compounds (amphetamine, dopamine and clozapine) cluster, potentially reflective of unique mechanisms of action.


Pathways. This tab displays pathway annotations from the KEGG and Reactome databases that are integrated with genes in CTD to provide insights into molecular pathways that may be affected by chemicals, and possible mechanisms underlying environmentally influenced diseases (Fig. 10).6, 10 Pathways are associated with chemicals in CTD when their member genes have curated interactions with a chemical of interest. Pathways are displayed in order of significance, which is calculated by the hypergeometric distribution and adjusted for multiple testing using the Bonferroni method. The hypergeometric distribution is used to calculate the probability that the fraction of interacting genes annotated to the pathway is significantly higher than the fraction of all human genes in the genome that are annotated to that pathway. The enrichment score is the negative log(base 10)-transformed corrected p-value (i.e., the larger the score, the more significant the enrichment). Data presented include the pathway name, pathway identifier, enrichment score, the interacting genes for that pathway (annotated genes), the fraction of interacting genes that are annotated to the pathway (cluster frequency), and the fraction of genes in the genome annotated to the pathway (genome frequency). Consistent with results discussed above that tie chlorpyrifos to neurologically important mechanisms, the top three associated pathways (ranked by number of underlying genes) are “neuroactive ligand receptor interaction,” “synaptic transmission,” and “signaling by GPCR.” Other pathways that tie together the roles of chlorpyrifos-interacting genes and associated diseases include “pathways in cancer,” “metabolism of xenobiotics by cytochrome P450,” and “drug metabolism.”

Figure 10
Figure 10 :

Chemical Pathway enrichment page. Chemicals in CTD are associated with KEGG and Reactome pathways via common curated genes. Data are organized by the pathway name, pathway identifier, enrichment score, the interacting genes for that pathway (annotated genes), the fraction of interacting genes that are annotated to the pathway (cluster frequency), and the fraction of genes in the genome annotated to the pathway (genome frequency). Consistent with ChemComps, GO enrichment and disease data for chlorpyrifos, the top three ranked pathways for this compound are involved in neurological functions (red rectangle).


GO. This tab lists the GO annotations that are statistically enriched for the genes that interact with a chemical or any of its child terms (Fig. 11). The terms provide insight into the biological properties that may be affected by a particular chemical. GO terms are displayed in order of significance, which is calculated by the hypergeometric distribution and adjusted for multiple testing using the Bonferroni method. The hypergeometric distribution is used to calculate the probability that the fraction of interacting genes annotated to the GO term or its descendants is significantly higher than the fraction of all human genes in the genome that are annotated to that GO term or its descendants. The enrichment score is the negative log(base 10)-transformed corrected p-value (i.e., the larger the score, the more significant the enrichment). Data are organized by the GO ontology (BP: Biological process; CC: Cellular component; and MF: Molecular function), GO level, GO term, enrichment score, the interacting genes for each annotated GO term (annotated genes), the fraction of interacting genes that are annotated to the GO term or its descendants (cluster frequency), and the fraction of genes in the genome annotated to the GO term or its descendants (genome frequency). The most significant terms associated with chlorpyrifos provide functional insights into its role in signaling, response to stimuli and neurological processes (transmission of nerve impulse, synaptic transmission, neurological system process, synapse, excitatory extracellular ligand-gated ion).

Figure 11
Figure 11 :

Chemical GO enrichment page. Data are organized by the Ontology (BP: Biological Process; MF: Molecular Function; CC: Cellular Component, GO Level, GO Term, Enrichment Score, the interacting genes for each annotated GO term (Annotated Genes), the fraction of interacting genes that are annotated to the GO term or its descendants (Cluster Frequency) and the fraction of genes in the genome annotated to the GO term or its descendants (Genome Frequency). The most significant terms associated with chlorpyrifos provide functional insights into its role in signaling, responding to stimuli and neurological processes (transmission of nerve impulse, synaptic transmission, neurological system process, synapse, excitatory extracellular ligand-gated ion; red rectangles)


References. This tab provides a list of all of the references in CTD that are associated with a chemical of interest or its descendants. Cited genes and diseases are also noted.

Genes

CTD contains curated data for approximately 20,000 genes from over 300 different species. Genes are curated, stored and presented using the cross-species gene vocabulary (symbols, names and synonyms) from the National Center for Biotechnology Information (NCBI) Gene database.11 Each CTD gene page provides linked identifiers to the corresponding species-specific pages in the Entrez Gene database.11 Genes that are believed to be unique to an organism or related group of organisms occupy separate pages. For example, the aryl hydrocarbon receptor (AHR) page includes AHR data for diverse species, excluding zebrafish, which is known to have several AhR paralogs with different nomenclature that occupy separate pages in CTD (Fig. 12a). Similar to the layout for CTD chemical pages, the Basic gene page provides nomenclature information and the top interacting chemicals. This page provides access to additional information via tabs for “Interactions,” “Chemicals,” “Diseases,” “GeneComps,” “Pathways,” “GO,” “References” and “Links.” The content of these pages is compiled in a manner analogous to what was described above for chemicals. Comparable genes presented on the GeneComps page are determined by the similarity index for interacting chemicals (Fig. 12b).

Figure 12a
Figure 12a :

(a) Basic gene information page. Information on Basic gene pages includes official nomenclature and synonyms, top curated interacting chemicals, and the NCBI Gene identifiers for the genes from vertebrates and invertebrates that are combined under this gene term. The nomenclature displayed corresponds to the human gene where available. The AHR-interacting chemicals include many that are well-studied environmental ligands of this receptor.


Figure 12b
Figure 12b :

(b) GeneComps. GeneComps provide the 20 genes with the most comparable set of interacting chemicals as the gene subject of the page. Similarity (or comparability) is measured by the Jaccard index, which is calculated as the size of the intersection of interacting chemicals for gene A and gene B divided by the size of the union of those chemicals. Overlapping chemical sets are provided (red arrows). Included here are many genes that are directly activated by ligand-bound AHR or function in the AHR signaling pathway.


Diseases

CTD contains curated data for over 4,000 diseases. Diseases are curated, stored and presented using official gene nomenclature from MeSH and the Online Mendelian Inheritance of Man (OMIM) database (http://www.ncbi.nlm.nih.gov/omim), which we merged by mapping OMIM terms to MeSH disease terms.2 Similar to chemicals and genes, detail pages for CTD diseases have data pages for: “Basic information,” “Interactions,” “Chemicals,” “Genes,” “Pathways” and “References”. The content of these pages is compiled in a manner analogous to what was described above for chemicals. Chemical and Gene pages for diseases display curated and inferred relationships (Fig. 13).

Figure 13
Figure 13 :

Disease Chemical page. Disease pages include curated and inferred gene and chemical relationships to provide insight into the potential influences on and mechanisms underlying disease etiologies. Shown here is a chemical page for Parkinson's disease in which the data were sorted by inference score. Rotenone and paraquat have both curated and inferred associations with this disease and novel information about genes that may help to explain these connections.


Tools

To help navigate the chemical gene disease data in CTD, we created a suite of analysis and visualization tools, accessible from the “Tools” tab. These include the following:

Batch Query. The Batch Query form allows users to download customized datasets associated with a list of chemicals, diseases, genes, GO terms or pathways of interest.2

VennViewer. The VennViewer tool allows users to compare associated datasets for up to three chemicals, diseases or genes (Fig. 14a). Options for associated datasets include chemical, gene, disease, pathway and GO (all or enriched) associations. An analysis of common inferred pathways for chlorpyrifos, Diazinon and clozapine indicates that there are 133 pathways common to all compounds including neurologically relevant processes (Fig. 14b). As expected, the two organophosphate insecticides (chlorpyrifos and Diazinon) have few unique associated pathways when compared to each other (one and three, respectively), whereas clozapine has 78 unique pathways.

Figure 14a
Figure 14a :

(a) VennViewer query form. Users may compare associated data for up to three chemicals, genes or diseases (top red arrow). Data to be compared may include curated interacting chemicals or genes, curated or inferred diseases, pathways or GO annotations (bottom red arrow).


Figure 14b
Figure 14b :

(b) VennViewer results. VennViewer results are presented graphically with hyperlinked data counts (within the diagram and below in the contents section; red rectangle) that allow users to identify the data that are unique or common to the input data. The Venn diagram shown here underscores the significant overlap in pathways shared by chlorpyrifos, Diazinon and clozapine (133 pathways), as well as a number of clozapine-specific pathways (78 pathways) that may reflect chemical-specific distinctions in its modes of action.


MyGeneVenn. The MyGeneVenn tool allows users to compare a gene list of interest to genes associated with up to two chemicals or diseases in CTD (Fig. 15a).3, 12 For example, genes that are significantly differentially regulated in response to a chemical exposure, as assessed by microarray analysis, can be compared to genes with curated interactions for that or other chemicals in CTD for the purpose of validation. Here significant gene interactions that were shown to be associated with exposure to tetrachlorodibenzo-p-dioxin (TCDD) during early zebrafish development are largely corroborated (90%) by curated TCDD gene interactions in CTD (Fig. 15b).12

Figure 15a
Figure 15a :

(a) MyGeneVenn query form. MyGeneVenn allows users to enter and compare a gene list of interest to genes associated with up to two chemicals or diseases in CTD (red arrows). This tool can be used to validate the experimental list of genes affected by a chemical or to explore similarity among potential modes of action or exposure conditions. Genes that were shown to be significantly affected by exposure to tetrachlorodibenzo-p-dioxin (TCDD) during early zebrafish development are entered for comparison with genes in CTD that have curated TCDD interactions.12


Figure 15b
Figure 15b :

(b) MyGeneVenn results. MyGeneVenn results are presented as a Venn diagram with hyperlinked data counts (within the diagram and below in the contents section; red rectangle) that allow users to identify the data that are unique or common to the input data. In the diagram shown here, 90% of the genes in the user-submitted data set have curated TCDD interactions in CTD.12


Conclusions and Future Directions

Improving understanding about the effects of the environment on human health is the major goal of CTD. To do this, we curate and integrate chemical gene interactions and chemical disease and gene disease relationships with other select data to inform hypothesis development about the mechanisms of chemical actions and potential etiologies of environmentally influenced diseases. Future development of CTD will expand the data scope and analysis capabilities to augment the value and interpretation of environmental health information. For example, planned developments include: (i) curation and integration of exposure data with chemicals, genes and diseases in CTD; (ii) new applications for existing data analysis measures such as DiseaseComps and pathway enrichment results for diseases; (iii) implementation of more sophisticated statistics for inferred data relationships; and (iv) addition of more graphical data visualization capabilities.

Carolyn J. Mattingly1

1. Mount Desert Island Biological Laboratory, Old Bar Harbor Rd, Salisbury Cove, Maine 04672, USA.
* Address correspondence to: cmattin@mdibl.org

References

  1. Davis, A.P., Murphy, C.G., Rosenstein, M.C., Wiegers, T.C. & Mattingly, C.J. The Comparative Toxicogenomics Database facilitates identification and understanding of chemical-gene-disease associations: arsenic as a case study. BMC Med. Genomics 1, 48 (2008).
  2. Davis, A.P. et al. Comparative Toxicogenomics Database: a knowledgebase and discovery tool for chemical-gene-disease networks. Nucleic Acids Res. 37, D786–D792 (2009).
  3. Davis, A.P. et al. The Comparative Toxicogenomics Database: update 2011. Nucleic Acids Res. 39, D1067–D1072 (2011).
  4. The Gene Ontology Consortium. Gene ontology: tool for the unification of biology. Nat. Genet. 25, 25–29 (2000). | Article | PubMed | ISI | ChemPort |
  5. Kanehisa, M. et al. KEGG for linking genomes to life and the environment. Nucleic Acids Res. 36, D480–D484 (2008). | Article | PubMed | ISI | ChemPort |
  6. Matthews, L. et al. Reactome knowledgebase of human biological pathways and processes. Nucleic Acids Res. 37, D619–D622 (2009). | Article | PubMed | ChemPort |
  7. Sayers, E.W. et al. Database resources of the National Center for Biotechnology Information. Nucleic Acids Res. 38, D5–D16 (2010).
  8. Davis, A.P. et al. GeneComps and ChemComps: a new CTD metric to identify genes and chemicals with shared toxicogenomic profiles. Bioinformation 4, 173–174 (2009).
  9. Barabasi, A.L. & Albert, R. Emergence of scaling in random networks. Science 286, 509–512 (1999). | Article | PubMed | ISI |
  10. Kanehisa, M., Goto, S., Furumichi, M., Tanabe, M. & Hirakawa, M. KEGG for representation and analysis of molecular networks involving diseases and drugs. Nucleic Acids Res. 38, D355–D360 , (2010). | Article | PubMed | ISI | ChemPort |
  11. Maglott, D., Ostell, J., Pruitt, K.D. & Tatusova, T. Entrez Gene: gene-centered information at NCBI. Nucleic Acids Res. 39, D52–D57 (2011).
  12. Planchart, A. & Mattingly, C.J. 2,3,7,8-TCDD upregulates FoxQ1b in zebrafish jaw primordium. Chem. Res. Toxicol. 23, 480–487 (2010).