Graphic: Courtesy of PharmGKB
Why does the cold medication that makes you sleepy give your friend the jitters? Diet, perhaps, or gender, but equally likely are your respective genetic backgrounds. With the era of personalized medicine approaching, individual responses to drugs are set to capture ever more attention from scientists and practitioners, and as a harbinger of the trend, the world's first public database of gene-drug interactions is now open for business.
Known as PharmGKB, for the Pharmacogenetics and Pharmacogenomics Knowledge Base, the resource is under the aegis of bioinformatician Russ Altman at Stanford University, who proclaims, "We want to be a well-known source of quality information about genetic variations in populations and their associations with drug-response phenotypes."
GENETIC PREDISPOSITIONS Pre-prescription genotyping is already mandatory for some drugs. In pediatric patients with acute lymphocytic leukemia, 5% will experience toxic effects when given a normal dose of 6-mercaptopurine, according to Mark Ratain at the University of Chicago.1 The unintended overdosing occurs because gene variations code for thiopurine methyltransferase, and the resulting altered forms of the protein slow down the drug's metabolic rate.
Another example of a genetic predisposition to adverse drug response is the connection between Bactrim (trimethoprim/ sulfamethoxazole) and long-QT syndrome, a potentially fatal cardiac arrhythmia. In patients taking the antibiotic, variant gene forms that encode a subunit of the cardiac potassium channel can result in impairment of the membrane depolarization cycle of the cardiac muscle. The otherwise clinically silent gene variation puts perhaps 1-2% of the general population at risk of drug-induced arrhythmia, according to a multicenter study.2
Similarly, variations in one of the genes coding for cytochrome P450, the drug metabolizing enzymes, can block the metabolism of codeine to morphine in 5-10% of the US population. "Typically this is not a big deal for patients taking codeine, because they just try another pain reliever if the codeine doesn't work," explains Altman, but the variability in response does show how hit-or-miss drug prescription can be today.
The most common DNA variants are called polymorphisms, generally defined as having greater than 1% prevalence in a given population. In the human genome there are perhaps 3 million polymorphisms, most occurring at single nucleotide sites, and thus designated as single nucleotide polymorphisms. Some SNPs may have detectable consequences, that is, drug or other interactions. Less common are insertions and deletions of larger tracts of DNA, known as mutations if they occur in coding or regulatory regions; they are likely to have adverse consequences for the proteins they code. Getting a handle on the complete picture of variability in drug action means tabulating the interactions between many drugs and the dizzying numbers of gene variants--this is the task of the PharmGKB database.
Photo: Courtesy of Russ Altman
COMPILING GENE-DRUG INTERACTIONS Two years ago a gut feeling about unknown gene-drug interactions led Altman and pharmacology scientists at about a dozen academic centers to form the Pharmacogenetics Research Network. Supported by $12 million from the National Institutes of Health, their studies include the gene-drug effects associated with asthma, cardiac problems, and cancer; the roles of genetic variability in drug response in ethnic populations; genetic differences and estrogen receptors; and the effects of gene variability on membrane transporters, which interact with one-third of all prescription drugs.
While the pharmacologists have been busy toting up evidence of the interactions, Altman and his team at Stanford have been buffing the database. It is practically empty now with just 128 submissions, mostly from Altman's own group, but he is hoping for 500 by mid-October and 1,000 by year's end. At Vanderbilt University in Nashville, Tenn., Dan Roden, an author of the cardiac arrhythmia study, is on the verge of sending in his data, while an appeal is being made to the pharmacology community through a "community-based" submissions mechanism linked to the site's main page at www.pharmgkb.org. One group of participants includes PharmD and PhD students in Mary Rellings' drug metabolism course at the University of Tennessee, Memphis. The students comb published literature, textbooks, and databases looking for evidence of interactions in genotyping, clinical outcomes, or pharmacokinetics studies.
What about the quality of the students' citations? Rellings says she'll check them before sending the submissions along to Stanford, but the main job of curating entries will be up to the Stanford group. Its first curator started work Oct. 1, and Altman says he expects to hire one or two more as soon as the rate of submissions picks up, possibly as soon as the end of the year.
The database's phenotype categories include clinical outcome, pharmacodynamics and drug response, pharmacokinetics, and molecular and functional assays. "This can be anything from a very lab-oriented phenotype like a binding constant of a drug, to a protein, all the way up to pupillary eye diameter or subjective sensation of pain," explains Altman.
FUTURE USERS Consumers of the new information will include pharmacogeneticists interested in the interaction of particular drugs with phenotype; another group, statisticians, are more broadly tackling the phenotype-genotype problem. "They want any dataset that relates any phenotype to any genotype," says Altman. "The fact that we are peddling drug phenotypes is fine with them, but that's not really what they care about. They are methods developers who want some good measure of human variation, and they'll use our datasets as practice."
Hoping for broad participation in the new resource, Altman looks at the success of GenBank, which took off once a policy of fast and early release of data was articulated, combined with journal publication contingent on depositing the sequence in GenBank. The similarity between GenBank and PharmGKB is limited, though, since the complexity of the latter's content and the difficult-to- anticipate nature of users' queries demands that the database schema, or underlying architecture, have a built-in flexibility that the relatively static format of GenBank does not require. The Stanford group used Protégé, a set of Java-based tools from the Stanford Medical Informatics program for designing ontologies, or blueprints for large knowledge systems like PharmGKB.3
Five to 10 years from now it's expected that information systems like PharmGKB will support point-of-care medicine. Then, genotype in hand, you'll know whether to take that cold remedy before you go to bed.
Potter Wickware is a freelance science writer in Mill Valley, Calif.
1. F. Innocenti, M.J. Ratain, "Update on pharmacogenetics in cancer chemotherapy," European Journal of Cancer, 38:639-44, 2002.
2. F. Sesti et al., "A common polymorphism associated with antibiotic-induced cardiac arrhythmia," Proceedings of the National Academy of Sciences, 97:10613-8, 2000.
3. D.L. Rubin et al., "Representing genetic sequence data for pharmacogenomics: an evolutionary approach using ontological and relational models," Bioinformatics, 18(suppl 1):S207-S215, 2000.