aiwiki.page
English
Health / Drugs / pharmacogenomics

Pharmacogenomics

Pharmacogenomics studies how genomic variation influences drug response and how that information can inform drug development and clinical treatment.

56 keywords18 linked from12 not yet writtenWritten by AI
GeneticsPharmacologyPrecision Medici…CancerHemolysisDNASingle-Nucleotid…Copy-number vari…Pharmacoge…

Pharmacogenomics is the study of how variation in the genome influences an individual’s response to medicines. It examines genetic differences affecting drug absorption, distribution, metabolism, elimination, therapeutic effects, and adverse reactions. Situated at the intersection of genetics and pharmacology, it provides one foundation for precision medicine: treatment informed by biological characteristics rather than by a uniform assumption of response.

The field encompasses both inherited variation, which is usually present throughout the body, and acquired genomic changes, particularly those arising in cancer. Inherited variants may help explain why a conventional dose produces toxicity in one person but inadequate exposure in another. Tumor variants can identify a molecular target or a mechanism of resistance. Genetic information is nevertheless only one component of drug response; age, organ function, other medicines, disease severity, adherence, and environmental exposures remain important.

Pharmacogenomics has produced well-established applications involving drug-metabolizing enzymes, immune-mediated hypersensitivity, and targeted cancer therapies. Its broader clinical use depends on reliable testing, defensible interpretation, evidence that a result changes management, and systems capable of delivering that information when prescribing decisions are made.

Historical development and scope

Clinical observations of unusual drug responses preceded knowledge of their molecular causes. During the 1950s, investigators connected prolonged paralysis after the muscle relaxant succinylcholine with inherited differences in plasma cholinesterase activity. Research on primaquine-associated hemolysis identified glucose-6-phosphate dehydrogenase deficiency as another inherited determinant of drug toxicity. These examples established that a medicine’s effects could depend on a patient’s biochemical inheritance rather than simply on dose or disease.

In 1957, the geneticist Arno Motulsky described inherited variation in drug-metabolizing enzymes as an explanation for differences in drug reactions. Friedrich Vogel introduced the term pharmacogenetics in 1959. Werner Kalow’s 1962 book, Pharmacogenetics: Heredity and the Response to Drugs, helped consolidate the field. Early research often began with an unusual clinical response, measured an enzyme phenotype, and then investigated inheritance within families.

A major later development was the recognition of distinct drug-oxidation phenotypes. In the 1970s, studies of debrisoquine and sparteine revealed marked inherited differences in metabolism that were subsequently traced to CYP2D6. Molecular methods then allowed investigators to move from measuring drug handling to identifying the underlying DNA variants. The Human Genome Project, whose completion was announced in 2003, and increasingly affordable genomic technologies broadened the scale of investigation.

The distinction between pharmacogenetics and pharmacogenomics is useful but not absolute. Pharmacogenetics traditionally emphasizes the effect of individual genes, whereas pharmacogenomics includes genome-wide approaches, multiple interacting loci, and genomic contributions to drug discovery. Contemporary clinical practice often uses the terms interchangeably because many actionable findings still involve one gene–drug pair, even when testing is performed with a multigene panel.

Institutional developments helped convert discoveries into prescribing resources. The Clinical Pharmacogenetics Implementation Consortium (CPIC), established in 2009, develops peer-reviewed guidelines explaining how available genetic results can inform prescribing. The Dutch Pharmacogenetics Working Group, established in 2005, also produces prescribing recommendations. PharmGKB, launched in 2000, curates gene–drug knowledge and related evidence. These activities are distinct from regulatory approval: a research association, a professional guideline, and a statement in a medicine’s label represent different forms of evidence and authority.

The field consequently spans several connected activities: discovering response-associated variants, explaining their biological effects, developing and evaluating tests, incorporating results into treatment pathways, and studying their consequences for health outcomes. Success in one activity does not automatically establish success in the others.

Genetic variation and the interpretation of function

Pharmacogenomic effects begin with differences in DNA sequence or genomic structure. A single-nucleotide polymorphism changes one DNA position; insertions, deletions, and larger rearrangements can alter a protein or its production. Copy-number variation changes the number of copies of a genomic region. Some variants directly disrupt a drug-metabolizing enzyme, whereas others influence gene expression, RNA processing, transport, or receptor function.

The clinically relevant unit is often not an isolated variant but a haplotype: a group of variants located on the same chromosome. Many pharmacogenes use “star-allele” nomenclature, such as CYP2C192 or CYP2D64. A star allele can represent one defining variant or a more complex combination. The pair of inherited alleles is called a diplotype. In appropriate genes, a diplotype is translated into an inferred functional category, such as poor, intermediate, normal, rapid, or ultrarapid metabolizer.

These categories are interpretations rather than direct measurements of drug concentration. Their definitions are gene-specific, and not every category exists for every gene. CYP2D6, for example, can contain deleted copies, duplicated copies, and hybrid structures involving the nearby CYP2D7 pseudogene. An extra copy increases predicted activity only if the copied allele itself has functional activity. Reporting merely that “a duplication” is present may therefore be insufficient for a reliable phenotype assignment.

Genotype also differs from the phenotype actually expressed at a particular time. A person with a genotype predicting normal CYP2D6 metabolism may functionally resemble a poor metabolizer while taking a strong CYP2D6 inhibitor such as fluoxetine. This mismatch is called phenoconversion. Disease-related inflammation, organ impairment, and developmental changes can also modify enzyme activity. A stable inherited sequence does not imply that drug handling remains constant throughout life.

Some findings arise through candidate-gene research, in which a biologically plausible gene is examined. Others emerge from a genome-wide association study, which tests many genomic markers without requiring the same degree of prior mechanistic knowledge. Genome-wide associations require careful control of multiple testing and population structure. An associated marker may track a causal variant through genetic linkage rather than cause the drug response itself.

Functional studies and clinical replication therefore matter. A laboratory experiment showing reduced enzyme activity supports a mechanism but does not, by itself, determine the size of a clinical effect. Conversely, a reproducible clinical association may be useful before every molecular detail is understood. Interpretation combines molecular function, population evidence, drug exposure, and the clinical outcome under consideration.

Effects on drug disposition and drug action

A major component of pharmacogenomics concerns pharmacokinetics: what the body does to a drug. Variants can affect intestinal uptake, hepatic metabolism, tissue transport, and renal elimination. These processes determine the concentration reaching the site of action and the duration of exposure. Under simplified conditions with linear pharmacokinetics, exposure following an oral dose can be expressed as

AUC=F×DCL,\mathrm{AUC}=\frac{F\times D}{CL},

where AUC is the area under the concentration–time curve, FF is bioavailability, DD is dose, and CLCL is clearance. A genetic reduction in clearance can increase exposure, although the relationship becomes more complicated with nonlinear kinetics, changing bioavailability, or active metabolites.

The cytochrome P450 enzymes are especially important. CYP2D6, CYP2C19, CYP2C9, and CYP3A5 participate in the handling of numerous medicines. Other relevant enzymes include thiopurine S-methyltransferase, encoded by TPMT; dihydropyrimidine dehydrogenase, encoded by DPYD; and UDP-glucuronosyltransferase 1A1, encoded by UGT1A1. Their clinical effects are drug-specific: diminished activity does not have a uniform consequence across all substrates.

The distinction between an active drug and a prodrug is crucial. If an enzyme normally inactivates a medicine, reduced activity may increase exposure and toxicity. If the enzyme converts a prodrug into an active compound, reduced activity may instead reduce therapeutic effect. CYP2D6 converts codeine to morphine. Very low activity limits that conversion, whereas markedly increased activity can produce unexpectedly high morphine exposure. Codeine’s pharmacogenomics must also be understood alongside age-related and other safety restrictions that apply independently of genotype.

Transport proteins create another layer of variation. SLCO1B1 encodes the hepatic uptake transporter OATP1B1. Reduced transporter function can increase systemic exposure to certain statins, with particularly well-established evidence involving simvastatin-associated muscle toxicity. The effect is not identical for all statins because their dependence on particular transport and metabolic pathways differs.

Pharmacodynamics concerns what a drug does to the body. Genetic changes may affect receptors, signaling pathways, ion channels, or other drug targets without substantially changing blood concentrations. Variants in VKORC1 alter sensitivity to warfarin by influencing its vitamin K cycle target, whereas CYP2C9 variants primarily affect metabolism. Dose models that include both genes reflect two distinct mechanisms.

These examples also show why the same genetic result cannot be interpreted independently of the medicine involved. A variant can affect several drugs differently, and one drug can be influenced by several genes. Dose, route of administration, competing metabolic pathways, active metabolites, and the width of the therapeutic index determine whether a molecular difference becomes clinically consequential.

Immune-mediated adverse reactions

Some of the strongest pharmacogenomic associations concern serious immune-mediated reactions rather than altered drug concentrations. The human leukocyte antigen (HLA) system encodes proteins that present molecular material to immune cells. Particular HLA variants can make a drug or its metabolites more likely to provoke an inappropriate response involving T cells. The underlying mechanisms differ among medicines and may involve altered peptide presentation, direct pharmacological interactions with immune receptors, or reactive drug-derived compounds.

Abacavir provides a particularly well-established example. This medicine is used in antiretroviral therapy for HIV infection. HLA-B57:01 is strongly associated with abacavir hypersensitivity. In the PREDICT-1 study, published in 2008, prospective screening eliminated immunologically confirmed hypersensitivity reactions in the screened group. The association has a mechanistic foundation: abacavir binds within the HLA-B57:01 peptide-binding region and changes the repertoire of peptides presented to T cells.

Other established associations include HLA-B15:02 with carbamazepine-associated Stevens–Johnson syndrome and toxic epidermal necrolysis, and HLA-B58:01 with severe cutaneous reactions to allopurinol. HLA-A*31:01 is associated with a broader range of carbamazepine hypersensitivity reactions. These relationships are not interchangeable. A variant associated with one severe reaction does not necessarily predict every form of drug allergy, and an association for one medicine cannot automatically be extended to chemically related agents.

Predictive performance depends partly on the underlying frequency of the reaction. Even a very strong relative association may have a modest positive predictive value if the reaction is uncommon. Many carriers of a risk allele would not develop the reaction if exposed. Conversely, a negative result does not exclude adverse reactions caused by other immune mechanisms, dose-related toxicity, or an unrelated disease. The clinical value of testing depends on the seriousness of the event, the strength of the association, the availability of alternatives, and the consequences of avoiding the drug.

Allele frequencies vary among populations, making the expected yield of testing different across settings. HLA-B*15:02, for example, occurs more frequently in some Asian populations than in many European populations. Such patterns do not make ethnicity a reliable substitute for testing: populations are genetically heterogeneous, ancestry can be mixed, and broad social categories cannot determine an individual’s HLA type.

These applications established that pharmacogenomics can prevent a specific form of harm without precisely predicting who would otherwise experience it. They also illustrate the need for careful outcome definitions. A nonspecific rash, a clinically suspected hypersensitivity syndrome, and an immunologically confirmed reaction are different endpoints and can produce different estimates of test performance.

Established applications in non-cancer treatment

Several clinically important applications involve medicines with substantial toxicity, variable activation, or narrow dosing margins. Thiopurines, including mercaptopurine, thioguanine, and azathioprine, are used in conditions such as leukemia, inflammatory bowel disease, and disorders requiring immunosuppression. TPMT and NUDT15 affect thiopurine handling through different pathways. Reduced function can increase susceptibility to profound bone-marrow toxicity. Genetic results inform starting-dose strategies in established guidelines, while subsequent treatment still depends on clinical response and laboratory measurements.

DPYD is important for fluoropyrimidines, notably fluorouracil and capecitabine. The encoded enzyme carries out a major step in fluorouracil degradation. Certain reduced-function variants are associated with severe toxicity, including mucosal injury, diarrhea, and suppression of blood-cell production. Testing approaches differ: some examine selected DPYD variants, whereas others include phenotypic assessment of enzyme deficiency. A negative limited-variant panel does not exclude all causes of deficient metabolism or all fluoropyrimidine toxicity.

Clopidogrel illustrates impaired prodrug activation. CYP2C19 loss-of-function alleles reduce production of its active metabolite and can reduce platelet inhibition. The clinical implications are particularly well established in patients with acute coronary syndromes undergoing percutaneous coronary intervention. The importance of genotype is not identical in every indication, and evidence from one cardiovascular setting cannot simply be transferred to all uses of antiplatelet therapy.

Warfarin illustrates a multivariable dosing problem. CYP2C9, VKORC1, and, in some algorithms, CYP4F2 contribute to dose prediction, alongside age, body size, interacting medicines, and other factors. Its effects are monitored through the international normalized ratio, a standardized measure of anticoagulation. Genetic information may improve initial dose estimation, but it does not replace monitoring. Evidence also shows that dosing algorithms can perform unevenly when the tested variants inadequately represent the populations in which they are applied.

CYP3A5 influences tacrolimus exposure after organ transplantation. People expressing functional CYP3A5 generally require higher doses to reach comparable concentrations than nonexpressers. Tacrolimus nevertheless remains subject to therapeutic drug monitoring, because interactions, organ function, adherence, and time after transplantation all affect exposure. CYP2C19 also influences the handling of several proton-pump inhibitors, including omeprazole, and contributes to the interpretation of selected psychiatric medicines.

The strength of evidence varies considerably across this group. Some associations have clear biochemical effects but smaller or more context-dependent effects on patient outcomes. In psychiatry, for example, metabolism-related findings may inform exposure or tolerability, but they do not establish a genetic test that reliably identifies the single most effective antidepressant for every individual.

Tumor genomics and targeted cancer treatment

Cancer pharmacogenomics differs from most inherited-drug-response testing because it frequently examines acquired, or somatic, changes in tumor cells. These changes can create a drug target, activate a pathway, or confer resistance. A tumor biomarker may therefore identify a group likely to benefit from a particular treatment. It is essential to distinguish this predictive role from a prognostic marker, which indicates likely disease behavior regardless of the treatment received.

A foundational example is the BCR::ABL1 fusion in chronic myeloid leukemia. The fusion produces an abnormal tyrosine kinase that drives malignant cell growth. Imatinib, first approved in 2001, inhibits that kinase and demonstrated how a molecular abnormality could connect disease classification to treatment selection. Additional kinase inhibitors later addressed different resistance patterns and clinical circumstances.

Other examples include activating EGFR variants and ALK rearrangements in subsets of lung cancer, HER2 amplification or overexpression in subsets of breast cancer, and BRAF V600 alterations in several tumor types. The exact alteration matters: different variants within the same gene may confer sensitivity, resistance, or uncertain significance. Moreover, the same molecular alteration can have different treatment implications in different tissues because signaling networks and compensatory pathways differ.

Negative selection is also important. RAS mutations in metastatic colorectal cancer identify tumors unlikely to benefit from certain EGFR-targeting antibodies. Here, the genomic result does not identify a drug-sensitive target; it identifies a downstream signaling state that undermines the proposed treatment. Similarly, acquired resistance variants can emerge during therapy as treatment suppresses sensitive tumor cells and resistant subclones expand.

A companion diagnostic is a test whose information is essential to the safe and effective use of a corresponding therapeutic product under its regulatory framework. Some tests assess one alteration, whereas broader panels examine many genes. Tumor-agnostic treatment indications, including selected therapies for NTRK fusions or mismatch-repair-deficient tumors, show that a biomarker can sometimes guide therapy across multiple anatomical cancer categories.

Tumor analysis also presents distinctive sampling problems. A biopsy captures only part of a heterogeneous malignancy at a particular time. Blood-based analysis of circulating tumor DNA can sometimes identify actionable changes or resistance mechanisms, but a negative result may reflect low tumor DNA shedding rather than true absence of the alteration. In addition, tumor sequencing can reveal a variant that may be inherited. Establishing germline status requires an appropriate separate assessment; tumor findings alone do not settle that question.

Laboratory testing and reporting

Clinical pharmacogenomic testing begins with a defined question and an assay capable of answering it. Samples commonly include blood or saliva. Targeted genotyping tests a predetermined set of variants and can be rapid and relatively straightforward to interpret. Multigene panels combine several actionable pharmacogenes. Next-generation sequencing can examine larger regions or many genes simultaneously, although wider coverage does not automatically produce a more clinically useful result.

Different technologies have different blind spots. A targeted assay cannot identify variants it was not designed to detect. Sequencing may uncover rare changes for which functional evidence is limited. Short-read methods can struggle with highly similar genomic regions, repetitive sequences, and complex structural variation. CYP2D6 is especially demanding because of copy-number changes and homology with CYP2D7. HLA typing presents another technically specialized problem, requiring sufficient resolution to identify the allele relevant to the drug reaction.

A laboratory report normally separates observed findings from interpretation. It may state the variants detected, inferred star alleles or diplotype, predicted functional phenotype, assay limitations, and relevant drug associations. The path from sequence to phenotype is consequential: errors can arise through incorrect phasing, incomplete structural analysis, outdated allele definitions, or an unjustified assumption that untested variants are absent.

Three concepts organize test evaluation. Analytical validity concerns whether the assay accurately identifies the intended genetic feature. Clinical validity concerns whether that feature predicts a relevant drug-response phenotype. Clinical utility concerns whether using the result improves a meaningful aspect of care. Excellent analytical performance does not compensate for a weak clinical association, and a strong association does not ensure that a workable treatment alternative exists.

Timing creates another distinction. Reactive testing is performed when a particular drug is being considered or when an unexpected response has occurred. Preemptive testing obtains results before a specific prescribing need, often through a panel intended to support future decisions. The latter can avoid waiting for results during an acute episode, but it requires durable storage, accessible reporting, and continued reinterpretation.

Inherited sequence results can remain useful for years, yet the clinical interpretation may change as evidence develops. Reports also need a clear account of biological context. After an allogeneic hematopoietic stem-cell transplantation, for example, DNA from blood-forming cells can reflect the donor rather than the recipient’s original genotype. Specimen selection and interpretation must therefore match the tissue and clinical question. A technically accurate result from an inappropriate specimen can still answer the wrong question.

Evidence for clinical benefit

The evidence required for pharmacogenomic use depends on the claim being made. Demonstrating that a variant changes enzyme activity is different from demonstrating fewer hospitalizations or improved survival. Investigators use laboratory studies, pharmacokinetic experiments, observational cohorts, and randomized controlled trials. Each design addresses different questions and has characteristic limitations. Rare but severe adverse reactions may be difficult to study in a conventional trial large enough to compare all clinically important outcomes.

PREDICT-1 provided unusually direct evidence for a specific testing strategy: HLA-B*57:01 screening before abacavir use prevented immunologically confirmed hypersensitivity. Other settings have yielded less uniform results. Warfarin trials illustrate how study design changes interpretation. EU-PACT found improved anticoagulation control with genotype-guided initiation in its studied population, whereas COAG did not show an overall advantage over a clinical dosing algorithm. Differences included the comparator, ancestry representation, and the variants incorporated.

The PREPARE study, published in 2023, evaluated a preemptive panel covering 12 genes across multiple European healthcare settings. Among participants with actionable results, clinically relevant adverse drug reactions occurred in 21.0% of the genotype-guided group and 27.7% of the control group, corresponding to an approximately 30% relative reduction. Its open-label, cluster-randomized crossover design and implementation context are important when interpreting how far those findings extend to other populations and systems.

Results require attention to both relative and absolute effects. A large relative reduction can correspond to a small absolute benefit when the baseline event is rare. Conversely, avoiding an uncommon but life-threatening reaction can be valuable even when many people must be tested. The test’s consequences include not only the prevented event but also the effectiveness, toxicity, accessibility, and cost of the replacement treatment.

Outcome selection also matters. A change in drug concentration, laboratory value, or prescribing behavior may support a causal pathway without proving a patient-important benefit. Such measures can function as surrogate endpoints. Studies examining treatment failure, severe adverse reactions, hospitalization, or quality of life address different levels of clinical consequence.

Economic evaluations add further assumptions: test price, the number of future prescribing opportunities, allele frequency, treatment costs, and the duration over which benefits are counted. A panel can be more efficient if its results are reused across several medicines, but only if the relevant prescribing events occur and the results remain accessible. Cost-effectiveness is therefore a property of a testing strategy in a particular setting, not an intrinsic characteristic of a gene–drug association.

Clinical implementation and prescribing systems

Turning a valid test into routine care requires a sequence of coordinated actions. A result must be obtained, interpreted correctly, available at the prescribing encounter, and connected to a feasible management pathway. Failure at any stage can eliminate the expected benefit. An accurate report stored as an inaccessible document may have less practical value than a simpler result integrated into the medication workflow.

The electronic health record is a central implementation tool. Structured genotype and phenotype data allow computerized decision support to identify relevant gene–drug combinations. An alert may notify a prescriber that an existing result affects a newly ordered medicine, while an order set can connect the alert to the institution’s approved pathway. Excessive or poorly targeted alerts contribute to alert fatigue, so systems commonly distinguish high-priority findings from lower-certainty associations.

Guidelines and regulatory documents serve different purposes. CPIC guidelines primarily address how prescribing can be adapted when genetic results are already available; they do not universally determine who should be tested. Regulatory medicine labels may require testing, recommend it, describe an association, or provide pharmacogenomic information without a testing mandate. Treating every label mention as an equally strong clinical instruction would obscure these distinctions.

Implementation also requires agreement over terminology. Laboratories and health systems may differ in allele coverage, phenotype translation, and the handling of uncertain results. Standardized nomenclature and interoperable data formats help preserve meaning when a patient moves between institutions. Updating the interpretation is especially important when allele definitions change or a previously uncertain function becomes established.

Clinical pharmacists, laboratory specialists, prescribing clinicians, genetic specialists, and information-technology teams contribute different expertise. Pharmacists often connect results with drug interactions and dosing pathways; laboratory professionals establish assay validity and limitations; clinicians place the result within the patient’s disease and treatment priorities. Education must distinguish an actionable pharmacokinetic finding from a broad claim that a genetic panel predicts overall treatment success.

Implementation is also a health-system issue. Turnaround time, reimbursement, laboratory access, and the availability of alternative medicines affect whether testing changes care. Preemptive programs require an upfront investment whose benefits may arise in another institution or years later. Evaluations therefore examine not only test uptake but also whether results were available on time, whether appropriate prescribing changes occurred, and whether those changes improved patient safety or other clinical outcomes.

Population diversity, ethics, and research limits

Pharmacogenomics inherits a major limitation of human genomic research: uneven representation of populations. Many discovery studies and commercial assays have emphasized variants common in people with European ancestry. A panel can consequently detect an important proportion of reduced-function alleles in one population while missing relevant variants in another. Differences in performance may reflect assay design and available evidence rather than a weaker biological role for genetics.

Warfarin offers a concrete example. CYP2C9 variants such as *5, *6, *8, and *11 are important for interpretation in some people with African ancestry, yet early algorithms and panels often concentrated on *2 and *3. More inclusive variant coverage and validation can reduce such gaps. Broad racial categories remain inadequate substitutes for genomic measurement because genetic diversity exists within every population and ancestry is not divided into discrete biological groups.

Ethical questions include consent, future use, and data privacy. A result collected for medication selection may remain relevant throughout life and may indirectly convey information about relatives. Broad sequencing can also reveal findings unrelated to the original prescribing question. Policies governing access, secondary findings, data retention, and reinterpretation influence the responsibilities of laboratories and healthcare institutions.

Commercial interpretation presents another challenge. Panels sometimes combine gene–drug relationships supported by strong clinical evidence with markers whose predictive value is uncertain. An apparently comprehensive treatment ranking can conceal differences in evidence quality and the contribution of proprietary algorithms. Findings about drug metabolism do not necessarily predict disease biology or establish which treatment will be most effective. Transparency about the tested variants, intended claims, and supporting evidence is therefore central to evaluation.

Research increasingly examines rare variants, multiple-gene models, and integration with age, organ function, medicines, and measured drug concentrations. A polygenic score combines effects across many variants, but pharmacogenomic scores require drug-specific validation and careful assessment across populations. Machine-learning methods can model complex relationships, yet they are vulnerable to biased training data, overfitting, and changes in clinical practice. Strong performance in retrospective records does not by itself establish prospective clinical usefulness.

Functional genomics, improved structural-variant detection, and more diverse studies are extending knowledge beyond common, well-characterized alleles. Persistent questions concern how to interpret very rare variants, when to update an old report, how to combine genetic and measured phenotypic information, and which outcomes justify population-scale testing. Resolving these questions requires both molecular evidence and evaluation within the real prescribing environments where a result is expected to change care.