The Pharmacogenetics and Pharmacogenomics Knowledge Base (PharmGKB: http://www.pharmgkb.org) is devoted to disseminating primary data and knowledge in pharmacogenetics and pharmacogenomics. We are annotating the genes that are most important for drug response and present this information in the form of Very Important Pharmacogene (VIP) summaries, pathway diagrams, and curated literature. The PharmGKB currently contains information on over 500 drugs, 500 diseases, and 700 genes with genotyped variants. New features focus on capturing the phenotypic consequences of individual genetic variants. These features link variant genotypes to phenotypes, increase the breadth of pharmacogenomics literature curated, and visualize single-nucleotide polymorphisms on a gene’s three-dimensional protein structure.

The drug, gene, and disease summary pages are entry points to more detailed information. Entering a drug, gene, or disease in the search box displays a list of hits pertaining to the query, with the option to narrow the search to information with phenotype data, genotype data, and/or literature annotation. Because a text search can hit different types of objects (e.g., genes, drugs, diseases, or literature citations), each hit is labeled as to the type of information it contains. For example, searching for a drug produces a first hit that points to an individual drug page. To provide faster access to the user’s relevant interests, the drug page is structured through a tab system displayed under the drug name (Fig. 2). The “Overview” tab provides generic and trade names for the entered drug. Under the “Properties” tab is information about molecular weight, indications, contraindications, mechanisms of action, and other pharmacokinetics data. Phenotypic primary data from clinical and basic science studies involving the entered drug are summarized under the “Datasets” tab. The “Curated Publications” tab displays a list of related genes and diseases gathered from literature. The pharmacogenetics literature is curated by using the five categories of evidence (CO, PK, PD, FA, and GN) and standardized vocabularies of genes, drugs, and diseases. The COEs and standard vocabularies provide an indexing scheme and highlight drug-gene and drug-disease relationships easily to the user. Genes are cataloged according to the Human Genome Nomenclature Committee terms (HGNC; Wain et al., 2002), and diseases are categorized by using Medline’s “Medical Subject Headings” terms (MeSH terms; Lindberg et al., 1993). Genes and diseases are clickable, leading to similarly organized pages with the tab system for easy navigation. Interactions of genes and drugs are visualized in drug-centered pathways that contain more in-depth knowledge.

A primary mission of the PharmGKB has been to serve as a repository for primary genotype and phenotype data related to pharmacogenetics and pharmacogenomics research from members of the scientific community. There is an increasing interest in creating larger patient cohorts in order to detect smaller effects with greater statistical power. Therefore, PharmGKB has initiated efforts to encourage researchers to pool data (particularly clinical data) in order to create larger data sets for analysis and sharing. These data-sharing consortia then work with PharmGKB curators to focus on a specific pharmacagenomics question. The International Warfarin Pharmacogenetics Consortium (IWPC) is the first such consortium and focuses on the development of a globally relevant warfarin pharmacogenetics dosing algorithm as its first specific research question. Warfarin is an anticoagulant that is difficult to dose and is associated with a set of genes whose variants are known to affect the optimal dose for individual patients. As a result of this collaborative world-wide effort, a new dosing algorithm has been defined that appears to perform better than a clinical-only or fixed-dose approach (IWPC, submitted). Encouraged by the IWPC’s success, PharmGKB has helped form a second consortium, the International Tamoxifen Pharmacogenomics Consortium (ITPC), with a focus on collecting and sharing data sets relating to CYP2D6 genotypes and clinical outcomes in women receiving adjuvant tamoxifen therapy for early breast cancer to gain a better understanding of the relationships between CYP2D6 metabolism status and tamoxifen (http://www.pharmgkb.org/views/project.jsp?pId=63). Tamoxifen, a selective estrogen-receptor modulator, is an antihormonal treatment for breast cancer patients whose tumors express the estrogen receptors and/or the progesterone receptor. It is used in women with metastatic breast cancer and women at high risk of developing breast cancer as a prevention strategy (Jordan, 2008). The cytochrome P450 enzyme, CYP2D6, is one of the major contributors to the metabolism of tamoxifen, especially through its role in the formation of antiestrogenic metabolites endoxifen and 4-hydroxytamoxifen. There are discordant reports relating to the relationship between CYP2D6 genotype and clinical outcomes with tamoxifen. Contrary to initial studies, which found no relationship, more recent studies have demonstrated that genetic variation of CYP2D6 and inhibitors of the enzyme markedly reduce endoxifen plasma concentrations in tamoxifen-treated women. Those studies showed that impaired tamoxifen metabolism results in worse treatment outcomes (Goetz et al., 2007; Schroth et al., 2007). The ITPC is committed to aggregating data on CYP2D6 and other genetic variants involved in the metabolism and effects of tamoxifen with clinical outcomes to address this complex question. This effort will clarify the role of the CYP2D6 genotype as an independent predictor for the treatment outcome of women with breast cancer with tamoxifen treatment. Thus, a priori genetic assessment of CYP2D6 could result in a more favorable tamoxifen treatment regime.

The PharmGKB initially provided only human-curated and annotated literature on pharmacogenetics and pharmacogenomics. “Non-curated Publications” is a new feature consisting of pharmacogenetics literature collected through automated computer algorithms. In particular, MScanner is a supervised learning-based classifier that retrieves relevant literature automatically using journal titles, and Medical Subject Headings (MeSH; Poulter et al., 2008). If one clicks on “Non-curated Publications,” and then on the “Details” link position after the article title, followed by “Marked-up Text,” the screen displays sentences from automatically retrieved articles with drug names, variant names, genes and gene products, and disease names highlighted and color-coded for each category, using a method based upon Pharmspresso (Fig. 6; Garten and Altman, submitted). The combination of automated literature curation and marked-up text serves as an additional pipeline to provide users more rapid access to the literature. Of course, the computer algorithms can introduce noise, including incorrect annotations, and the curators work to review automated annotations and “upgrade” them to two stars (**).

Nonsynonymous polymorphisms in coding exons may occur that affect structure-function relationships. With the development of high-throughput sequencing technology, researchers are able to identify new SNPs before characterizing their functional relevance. Mapping coding SNPs onto 3D protein structures enables researchers to visualize the macromolecular context of the variants. Initial efforts at mapping coding SNPs onto 3D protein structures use X-ray and nuclear magnetic resonance (NMR) structures stored in the Protein Data Bank (RCSB PDB; Berman et al., 2003). PharmGKB uses the MBT Protein Workshop software from the PDB for structure viewing over the Web. Structures mapped with variant location, when available, can be accessed from the “Variants” tab. Annotated structures are displayed as they are curated. Figure 7 shows the PDB structure for angiotensin I converting enzyme (ACE) with the locations of coding SNPs highlighted in red.

Bridging the fields of pharmacology, genetics, and medicine, the PharmGKB provides diverse researchers access to pharmacogenomics research results and discoveries. VIP summaries, pathway diagrams, variant annotations, literature annotations, and other features provide high-level summaries of knowledge that can be used to design new studies. The PharmGKB can also be used as an educational tool to teach clinical and basic science students about pharmacogenetics (Owen et al., 2007). All primary data submitted to PharmGKB by researchers are available to view and download. Recently, PharmGKB has taken on an exciting new role as an independent broker for data-sharing consortia focused on answering important pharmacogenomics questions. Results of the first consortium focused on warfarin has yielded a warfarin-dosing algorithm, based on studies of thousands of patients (IWPC, submitted). A new consortium is addressing the pharmacogenetics of tamoxifen, thus continuing PharmGKB’s involvement in multicenter, international studies.