23andMe vs Clinical Genomics: Which DNA Results Should You Actually Trust?

You spit into a tube, send it off, and a few weeks later receive a report telling you your ancestry composition, your likelihood of having wet earwax, and — if you opted in — your estimated risk for Parkinson's disease or breast cancer. Consumer direct-to-consumer (DTC) genetic testing has made genomics accessible to tens of millions of people at a price point below a hundred dollars. Clinical genomic testing, meanwhile, costs thousands, requires a physician's order, and arrives with a genetic counselor's interpretation. The outputs look similar on the surface: both involve your DNA, both produce risk estimates, both generate reports with variant names. But the underlying science, regulatory standards, and clinical validity are worlds apart.

Understanding that difference is not merely an academic exercise. Patients who receive a negative consumer result for BRCA1 or BRCA2 variants sometimes conclude they have no elevated breast cancer risk — a potentially dangerous misreading of what that result actually means. Conversely, patients who receive a positive signal on a consumer test may undergo unnecessary anxiety or medical procedures without proper clinical confirmation. Getting genomics right matters enormously, and the first step is understanding what each testing modality actually measures, and what it does not.

What Consumer DTC Tests Actually Measure

Genotyping Arrays: A Selected Snapshot

Services like 23andMe use a technology called SNP genotyping arrays. A microarray chip contains hundreds of thousands of probes — typically between 600,000 and 700,000 — each designed to detect a specific, pre-defined single-nucleotide polymorphism (SNP). A SNP is a position in the genome where a single base differs between individuals. The chip interrogates only those pre-selected positions; everything else in the genome is invisible to it. The human genome contains approximately three billion base pairs and an estimated four to five million common variants in any individual. Genotyping arrays therefore capture a small, curated fraction of total genetic variation.

The variants selected for these arrays tend to be common — those with a minor allele frequency above one percent in reference populations. This design choice is deliberate: common variants are well-characterized, their population frequencies are established, and the statistical associations between them and disease traits are often documented in genome-wide association studies (GWAS). The tradeoff is that rare, highly penetrant variants — the ones most likely to cause serious hereditary conditions — are frequently absent from the chip entirely. This is not a flaw in execution; it is a fundamental constraint of the technology.

Regulatory Status of DTC Health Reports

In the United States, the FDA classifies DTC genetic health risk tests as Class II medical devices requiring premarket review. 23andMe received authorization for several categories including carrier status reports, pharmacogenomics reports, and health risk reports for conditions such as late-onset Alzheimer's disease, Parkinson's disease, hereditary thrombophilia, and BRCA-related cancer risk. FDA authorization means the test has demonstrated analytical validity (it reliably detects the variants it claims to detect) and some level of clinical validity (the detected variants are genuinely associated with the claimed condition). It does not mean the test has the same sensitivity or comprehensiveness as a clinical diagnostic test, nor does it establish clinical utility — the ability of the test result to improve patient outcomes.

Importantly, many features in consumer reports exist outside FDA oversight. Trait reports, wellness scores, and ancestry compositions are not FDA-regulated health claims. The scientific backing for these features varies enormously — some are well-supported, others are based on correlations that have not replicated across populations. This distinction matters when patients bring consumer results to a clinical encounter: physicians and genetic counselors must first assess which results carry any clinical weight and which are essentially entertainment.

What Clinical Genomic Testing Looks Like

From Targeted Panels to Whole-Genome Sequencing

Clinical genomics is not a single technology — it is a spectrum of tests ordered to answer a specific clinical question. At one end are single-gene tests, used when clinical presentation strongly suggests a particular condition (for example, sequencing only CFTR when cystic fibrosis is suspected). Multi-gene panels test a curated set of genes simultaneously — a hereditary breast and ovarian cancer panel might include BRCA1, BRCA2, PALB2, ATM, CHEK2, and a dozen other genes with established cancer risk associations. These panels are designed by clinical laboratories using current evidence and are updated as new gene-disease associations achieve sufficient clinical validity.

Whole-exome sequencing (WES) reads all protein-coding regions of the genome — roughly one to two percent of total DNA but the region where the majority of known disease-causing variants reside. Whole-genome sequencing (WGS) reads all three billion base pairs, capturing coding and non-coding regions, structural variants, copy number variations, and novel mutations that panels and exome sequencing miss. Clinical WGS is increasingly used for rare and undiagnosed disease evaluation, neonatal intensive care settings, and — as costs continue to fall — is beginning to appear in broader precision medicine programs.

Variant Interpretation: The Critical Difference

Finding a variant in a gene is only the beginning. The critical question is what that variant means biologically. Clinical laboratories apply the American College of Medical Genetics and Genomics (ACMG) variant classification framework, which categorizes variants as pathogenic, likely pathogenic, variant of uncertain significance (VUS), likely benign, or benign. This classification draws on population databases (gnomAD, ClinVar), functional studies, co-segregation data from families, computational predictions, and expert panel curation. The process requires trained molecular pathologists or variant scientists and is subject to laboratory accreditation standards (CLIA in the US, ISO 15189 internationally).

Consumer tests do not perform this type of rigorous variant classification for the full range of variants they might detect. They report on a pre-defined set of variants with established population-level associations, which is a different — and more limited — scientific exercise. A clinical report that says a variant is "pathogenic" carries a specific, regulated, evidence-based meaning. A consumer report that says your risk is "slightly elevated" is a statistical estimate derived from GWAS associations in a reference population and may not translate meaningfully to any individual's actual risk trajectory.

Polygenic Risk Scores: Promise and Pitfalls in Both Contexts

What Polygenic Scores Actually Represent

Both consumer and clinical genomics increasingly incorporate polygenic risk scores (PRS) — weighted sums of dozens to millions of variants that each contribute small effects to overall disease risk. PRS for conditions like coronary artery disease, type 2 diabetes, and certain cancers have demonstrated meaningful predictive power at the population level, with individuals in the highest percentile carrying risk several times greater than the population average. This is genuinely useful information when properly contextualized.

Consumer PRS products vary considerably in methodology and transparency. A score built on GWAS data from predominantly European ancestry populations performs less well when applied to individuals of South Asian, African, or admixed ancestry — sometimes dramatically so. This is not a minor statistical caveat; it means the same score can be well-calibrated for one person and systematically miscalibrated for another based purely on ancestry. Clinical PRS implementations are beginning to address this through ancestry-specific models and recalibration approaches, but the problem is not yet solved in either domain. Anyone using a polygenic risk score — consumer or clinical — should understand the ancestry composition of the training dataset and treat cross-ancestry extrapolation with appropriate caution.

When PRS Meets Clinical Decision-Making

Clinical integration of PRS is advancing most rapidly in cardiovascular medicine, where high polygenic risk for coronary artery disease can identify individuals who would benefit from earlier statin therapy or more aggressive risk factor management regardless of traditional risk calculators. In oncology, PRS is beginning to inform screening age recommendations — for example, adjusting when a woman should begin mammography based on her combined monogenic and polygenic breast cancer risk. These applications exist within regulated clinical frameworks with defined cutoffs, clinical guidelines, and physician oversight. Consumer PRS results, arriving without clinical context, can lead patients to under-react (because population-level risk percentages feel abstract) or over-react (by demanding investigations not indicated by guidelines) in equal measure.

Pharmacogenomics: Where the Gap Has Real Consequences

Drug Response Is Written in Your Genome

Pharmacogenomics — the study of how genetic variation affects drug metabolism and response — is one of the most clinically actionable branches of genomic medicine. Variants in genes like CYP2D6, CYP2C19, CYP2C9, SLCO1B1, and TPMT determine whether a patient is a poor metabolizer, normal metabolizer, rapid metabolizer, or ultrarapid metabolizer of hundreds of commonly prescribed medications. These classifications directly affect dosing, drug selection, and the risk of adverse events. Codeine can cause opioid toxicity in CYP2D6 ultrarapid metabolizers. Clopidogrel is ineffective in CYP2C19 poor metabolizers. Simvastatin causes muscle toxicity at elevated rates in SLCO1B1 risk allele carriers.

Consumer tests include some pharmacogenomic information, but clinical pharmacogenomic testing covers a broader gene panel with more comprehensive variant coverage within each gene, uses validated clinical decision support tools aligned with CPIC (Clinical Pharmacogenomics Implementation Consortium) guidelines, and integrates directly with electronic health records and prescribing workflows. The difference is not whether the underlying variant is real — a CYP2D6 poor metabolizer classification from 23andMe reflects genuine biology — but whether the result is complete, properly interpreted, and actionable within a care context. A patient managing multiple medications should seek clinical pharmacogenomic testing rather than relying on consumer reports that may miss functionally important variants or lack the clinical infrastructure to translate results into prescribing guidance.

Data Privacy: What Happens to Your DNA After the Test

Consumer Services and the Data Economy

When you send your DNA to a consumer testing company, you are doing more than purchasing a health report. You are entering a data relationship that may persist indefinitely. Consumer genetic data has become a significant commercial asset — useful for pharmaceutical research partnerships, population genetics studies, and potentially insurance and law enforcement contexts. Privacy policies vary by company and can change over time. The 23andMe bankruptcy proceedings that began in 2025 put into stark relief what happens to genetic databases when a company faces financial distress: customer genetic data became a contested asset in those proceedings, raising fundamental questions about the long-term security of consumer DNA deposits.

Clinical genomic data operates under different legal protections. In the United States, clinical genetic test results are protected health information under HIPAA. The Genetic Information Nondiscrimination Act (GINA) prohibits health insurers and employers from discriminating based on genetic information, though notably it does not cover life insurance, disability insurance, or long-term care insurance. Clinical laboratory data sharing for research requires informed consent processes with specific disclosures about data use. This regulatory framework is imperfect and still evolving, but it offers substantially stronger privacy architecture than the terms of service agreements governing consumer genetic data — and understanding who owns your health data is increasingly critical in an era of genomic medicine.

AI and the Future of Genomic Interpretation

Machine Learning Is Reshaping What Genomic Data Can Tell Us

The gap between consumer and clinical genomics is not static — it is being actively reshaped by artificial intelligence and machine learning applied to genomic data. Deep learning models trained on large genomic datasets are improving the interpretation of variants of uncertain significance, predicting splicing effects of non-coding variants, and identifying functional consequences of mutations that traditional bioinformatic pipelines miss. AI applied to genomics is also enabling more sophisticated polygenic risk models that better account for ancestry heterogeneity and gene-environment interactions.

Consumer companies are investing heavily in AI-driven phenotyping — using self-reported survey data combined with genotype data to discover new gene-trait associations at scale. The massive datasets these companies hold are genuinely scientifically valuable. At the same time, the clinical genomics space is deploying AI for automated variant classification, real-time literature integration, and phenotype-driven differential diagnosis in rare disease contexts. These developments are narrowing some gaps while creating new ones: as AI interpretation tools proliferate, the question of which interpretations are validated, reproducible, and clinically trustworthy becomes more complex, not less. The fundamental principle remains: an AI-generated interpretation applied to a genotyping array result carries different epistemic weight than an AI-assisted interpretation of clinical whole-genome sequencing performed under CLIA standards.

Where Precision Medicine Is Headed

The long-term trajectory of genomic medicine is toward integration — combining germline genomic data with other molecular layers including the transcriptome, proteome, metabolome, and microbiome into comprehensive individual health profiles. The microbiome, for instance, interacts with genetic variants to modulate drug metabolism, immune function, and disease risk in ways that neither test captures independently. Epigenetic modifications — which affect gene expression without altering the DNA sequence itself — represent another layer of biological information that neither genotyping arrays nor standard sequencing fully captures. As costs fall and multi-omic integration becomes computationally tractable, the binary of consumer versus clinical genomics may give way to a continuum of testing depth calibrated to clinical need. For now, however, the analytical and regulatory gap between the two remains wide enough to matter enormously in practice.

A Practical Guide: Which Test for Which Situation

When Consumer Testing Is Appropriate

Consumer DTC testing serves genuine purposes when expectations are properly calibrated. Ancestry composition, deep genealogical connections to relatives, and broad wellness trait information are areas where consumer tests provide value without clinical pretension. For individuals who are simply curious about their general genetic landscape and have no significant personal or family history of heritable disease, a consumer test can be an engaging entry point into genomics literacy. The key is understanding that this is exploration and education, not medical diagnosis. Results that generate curiosity — a possible elevated risk signal, an unexpected ancestry composition, a pharmacogenomic flag — are best treated as reasons to have a conversation with a physician rather than as conclusions in themselves.

When Clinical Testing Is Essential

Clinical genomic testing is appropriate — and often essential — in a range of situations. A personal or family history of hereditary cancer syndrome (breast, ovarian, colorectal, pancreatic, or others), hereditary cardiovascular disease (cardiomyopathy, arrhythmia syndromes, familial hypercholesterolemia), or neurological conditions with known genetic drivers all warrant clinical evaluation that goes far beyond consumer testing. Rare and undiagnosed disease workup, pediatric developmental delay or intellectual disability evaluation, and prenatal carrier screening in the context of family planning all require clinical-grade testing with physician interpretation and genetic counseling. Pharmacogenomic guidance for polypharmacy or when considering medications with narrow therapeutic windows should similarly be pursued through clinical channels. And whenever a consumer result — positive or negative — is being considered as the basis for a medical decision, clinical confirmation is necessary before that decision is made.