Whole Genome Sequencing and Polygenic Risk Scores: What Consumer Genomics Can Predict

The envelope that arrives in the mail looks like any other. Inside is a plastic tube, a barcode and an invitation to learn, from a few milliliters of saliva, what your DNA says about your heart, your memory and your children. As of January 2026 that invitation is landing in more mailboxes than ever, and the search traffic shows it: whole genome sequencing and the newer phrase polygenic risk score have both jumped in interest over the past year.

Two things lit the fuse. In June 2025 the health service in England set out an ambition to offer whole genome sequencing to every newborn within a decade, building on a 100,000-baby research study already under way. A few months earlier, the bankruptcy of a well-known consumer DNA-testing company sent millions of customers scrambling to find out who now owned their genetic data.

So the questions people are typing are practical ones. What does a full readout of the genome actually tell you? What is a polygenic score, and should a number on a dashboard change how you live? Here is the evidence, graded honestly.

Why whole genome sequencing is trending right now

Genomics has had false dawns before. The Human Genome Project finished its reference sequence in April 2003, and for years afterward the technology stayed inside research labs. What is different about this moment is scale: sequencing a complete genome, once a multi-year national effort, now takes a few days on a benchtop machine, and the price has fallen far enough that national health programs are discussing it for healthy babies, not just sick ones.

Three threads have braided together in search interest. The first is the newborn screening conversation. England’s Generation Study, launched in October 2024, is sequencing the genomes of 100,000 newborns to look for roughly 200 rare, treatable childhood conditions, and the government’s June 2025 ten-year plan floated extending that to every baby. Parents hear “every newborn” and start reading.

The second thread is privacy. When a consumer genomics company enters bankruptcy, its most valuable asset is often its database of customer DNA. That realization sent people back to the terms of service they had clicked past, and from there into searches about whether genetic data can be sold, subpoenaed or used by insurers.

The third is the polygenic risk score. For a decade these scores lived in academic papers. Now several consumer platforms show one for coronary artery disease, type 2 diabetes or breast cancer, often as a percentile bar. A percentile feels like a verdict. The science behind it is real but far more modest than the dashboard suggests, and that gap between presentation and evidence is what this piece is about.

What changed recently

Because this is a fast-moving field, it helps to pin the recent facts to dates.

• April 2003: the Human Genome Project declared the human reference sequence essentially complete, covering about 92 percent of the genome at high accuracy. The remaining stubborn stretches, mostly repetitive DNA near chromosome tips and centers, were not closed until a separate consortium published a gap-free sequence in 2022 (NIH National Human Genome Research Institute).
• 2021 onward: the American College of Medical Genetics and Genomics has steadily expanded its list of “secondary findings” genes, meaning genes a lab should report if it finds a harmful variant during sequencing ordered for another reason. The list now contains roughly 80 genes, almost all tied to conditions where early action changes outcomes, such as hereditary breast and ovarian cancer syndromes, Lynch syndrome and familial hypercholesterolemia.
• October 2024: England’s Generation Study began enrolling newborns for whole genome sequencing within a research framework, with results fed back to families and pediatricians (NHS).
• March 2025: a major direct-to-consumer DNA company filed for bankruptcy protection, prompting several state attorneys general to remind customers of their right to delete stored data.
• June 2025: the NHS ten-year plan for England described an aim to offer genome sequencing to all newborns within ten years, framed as a long-term ambition rather than a launched program.

What has not changed is just as telling. No major guideline body, including the US Preventive Services Task Force, currently recommends whole genome sequencing or polygenic risk scores as routine screening for healthy adults. The professional statements that exist treat polygenic scores as promising research tools whose clinical place is still being defined. Hold both facts at once: the technology is moving quickly, and the guidance is deliberately moving slowly.

What is whole genome sequencing, and how is it different from a drugstore DNA kit?

Whole genome sequencing is a laboratory method that reads the order of nearly all 3 billion DNA letters, called bases, in a person’s genome. That includes the roughly 20,000 protein-coding genes and the much larger stretch of non-coding DNA between them that helps switch those genes on and off.

Most consumer kits do something far narrower. They use a genotyping array, a chip that checks a fixed menu of several hundred thousand to a million pre-selected single-letter variants, known as SNPs (single nucleotide polymorphisms). Picture the genome as a 3-billion-character book. Sequencing reads every page. An array flips to about 700,000 bookmarked spots and records which letter sits there. If your family carries a rare spelling change that was never bookmarked, the array will simply never see it.

In between sits whole exome sequencing, which reads only the protein-coding portion, about 1 to 2 percent of the genome. Because most known disease-causing variants fall in that 1 to 2 percent, exome sequencing captures a large share of medically useful information at a fraction of the data volume. Whole genome sequencing adds the non-coding regions, catches some structural changes more evenly, and avoids the capture step that can leave gaps in exome data (MedlinePlus Genetics).

One more distinction matters for consumers. Clinical sequencing is ordered by a clinician, run in an accredited laboratory, interpreted by a medical geneticist and delivered with counseling. Consumer sequencing may use the same chemistry but typically returns a file and an app. The chemistry is not the hard part. Interpretation is.

What is the process of whole genome sequencing?

The journey from spit tube to report has five stages, and knowing them helps explain both the power and the blind spots.

Sample and DNA extraction. A saliva or blood sample contains white blood cells or cheek cells. The lab breaks them open and purifies the DNA, which arrives as extremely long strands.

Library preparation. Those strands are sheared into fragments, usually a few hundred bases long for the most common technology, and tagged with short synthetic adapters so the machine can grip them.

Sequencing. The instrument reads each fragment many times. Clinical labs typically aim for about 30-fold coverage, meaning each position in the genome is read, on average, 30 times. That redundancy lets software distinguish a real variant from a machine error. Newer long-read platforms read fragments tens of thousands of bases long, which helps in repetitive regions, though they are less common in consumer products.

Alignment and variant calling. Software lines the fragments up against the human reference genome and lists every place where the sample differs. A typical person differs from the reference at 4 to 5 million positions.

Interpretation. This is where the time goes. Of those millions of differences, nearly all are harmless. Analysts filter for variants in known disease genes, compare them against curated databases, and classify each as benign, likely benign, uncertain, likely pathogenic or pathogenic. A clinical report usually names a handful of findings. The raw file behind it runs to about 100 gigabytes.

Turnaround for a clinical test is often several weeks. Rapid sequencing for critically ill newborns, where a diagnosis can change treatment in intensive care, can be completed in days, and that use case has some of the strongest evidence in the entire field.

What were the two methodologies of the Human Genome Project?

This question shows up in search because it appears on biology exams, but the answer also explains why modern sequencing works the way it does.

The publicly funded Human Genome Project used a hierarchical shotgun approach, sometimes called clone-by-clone. Scientists first broke the genome into large overlapping chunks of roughly 150,000 bases, mapped where each chunk belonged on its chromosome, and only then shattered each chunk into small pieces for sequencing. Because every small piece was already known to come from a specific mapped region, reassembly was like solving a jigsaw puzzle one section at a time.

A private company, racing the public effort in the late 1990s, used whole-genome shotgun sequencing. It skipped the mapping step entirely, fragmented the whole genome at once, sequenced everything, and relied on computing power to find overlaps and stitch the pieces back together. Critics worried that repetitive DNA, where the same sequence appears thousands of times, would make the puzzle unsolvable. The approach worked well enough that both groups announced draft sequences together in June 2000, and the public project’s finished sequence followed in 2003 (NIH National Human Genome Research Institute).

Whole-genome shotgun is the ancestor of every sequencer in use today. The reason it now succeeds where early critics expected failure is that we have a high-quality reference genome to align against. Instead of assembling a puzzle from scratch, a modern lab lays each fragment onto a finished picture and notes where the colors differ. That shortcut is also why repetitive regions remain the weakest part of standard short-read sequencing, a limitation that returns in the section on what the test cannot detect.

What is a polygenic risk score, in plain terms?

A polygenic risk score is a single number that adds up the small effects of many common genetic variants to estimate how a person’s inherited risk for a disease compares with the rest of a population. The word polygenic means many genes. For common conditions such as coronary artery disease, type 2 diabetes or most breast cancer, there is no single gene that decides the matter; there are thousands of variants, each nudging risk up or down by a fraction of a percent.

The scores come from genome-wide association studies, or GWAS, which compare the DNA of hundreds of thousands of people with and without a condition to find variants that appear more often in the affected group (MedlinePlus Genetics). Researchers assign each variant a weight, multiply by how many copies a person carries, and sum the result. The output is usually expressed as a percentile: a score in the 95th percentile means higher inherited risk than 95 percent of the reference population.

Two features of that definition carry most of the weight in the sections that follow. First, the score is relative. Being in the top 5 percent for a rare disease can still mean a low absolute risk. Second, the score is only as good as the reference population it was built from, and the vast majority of GWAS participants to date have been of European ancestry (NIH National Human Genome Research Institute).

Think of a polygenic score as a weather forecast built from historical patterns rather than a diagnosis. A forecast of 70 percent rain is genuinely useful. It still tells you nothing certain about whether your particular street gets wet.

What can whole genome sequencing predict with real confidence?

Where sequencing earns its reputation is in single-gene, or monogenic, conditions: disorders where one damaged gene is enough to cause disease, or to raise lifetime risk dramatically. Here the evidence is mature and the clinical pathways are established.

Across large biobank studies, roughly 2 to 4 percent of apparently healthy adults carry a pathogenic variant in one of the actionable genes on the American College of Medical Genetics and Genomics list. The most common findings fall into three groups: hereditary cancer syndromes involving genes such as BRCA1, BRCA2 and the Lynch syndrome mismatch-repair genes; inherited heart conditions including cardiomyopathies and arrhythmia syndromes; and familial hypercholesterolemia, a genetic cause of very high LDL cholesterol that affects about 1 in 250 people and is often undiagnosed.

For each of those groups, finding the variant changes what happens next in ways supported by guidelines: earlier or more frequent screening, specialist follow-up and, in some cases, preventive treatment decided with a clinician. Crucially, that benefit shows up in observational cohorts and in the clinical experience of genetics services rather than in randomized trials of population sequencing, which have not yet been completed.

Sequencing also reliably identifies carrier status for recessive conditions such as cystic fibrosis, which matters for family planning, and a growing set of pharmacogenomic variants that affect how the body handles certain medicines. Whether a pharmacogenomic result should alter a prescription is a conversation for the prescribing clinician, not something to act on from a consumer report.

Notice what is not on this list. Everyday risk for heart disease in someone without a monogenic variant, risk of Alzheimer’s disease beyond the well-known APOE association, longevity, athletic talent and intelligence are all areas where sequencing can produce a number but the number has limited power to predict an individual life.

What can whole genome sequencing not detect?

A full genome sounds like a complete picture. It is not, for reasons that are partly technical and partly biological.

Repetitive regions and repeat expansions. Standard short-read sequencing struggles wherever the same short sequence repeats many times. Conditions caused by expanded repeats, such as Huntington disease and fragile X syndrome, are often confirmed with dedicated tests rather than a routine genome readout.

Genes with near-identical copies. Some medically important genes sit beside a pseudogene, a non-functional twin with almost the same spelling. Short reads can land on the wrong copy. The gene behind spinal muscular atrophy and several drug-metabolizing genes fall into this category and usually require specialized analysis.

Low-level mosaicism. If a variant is present in only a small fraction of cells, 30-fold coverage may miss it entirely.

Epigenetics. Chemical tags on DNA, such as methylation, switch genes on and off without changing the sequence. Standard sequencing does not read them, yet they shape how genes behave in response to age, diet and environment.

Variants of uncertain significance. Sequencing detects the change but cannot always say what it means. Roughly a third of people who undergo broad sequencing receive at least one variant of uncertain significance, and most are later reclassified as benign.

Everything that is not DNA. Smoking, blood pressure, air quality, physical activity and plain chance account for most of the variation in who develops common diseases. For coronary artery disease, for example, a polygenic score and traditional risk factors each capture a piece of risk, and neither captures the whole.

A genome is a parts list for a human being. It is not the maintenance record, and it cannot see the road ahead.

What the evidence actually says, graded by strength

Different claims in consumer genomics rest on very different foundations. Separating them is the single most useful thing a reader can do.

Strong evidence (large observational cohorts, consistent across populations, embedded in guidelines): sequencing identifies monogenic variants with high analytic accuracy, and for actionable genes the resulting care changes are supported by established clinical pathways. Rapid genome sequencing in critically ill infants has shown diagnostic yields of roughly 30 to 50 percent in several cohort studies, with documented changes in management.

Moderate evidence (large biobank studies, limited randomized data): polygenic risk scores stratify populations. In the most cited coronary artery disease study, people in the top 8 percent of the score had about three times the risk of those in the middle, comparable to carrying a monogenic familial hypercholesterolemia variant. For breast cancer, women in the top 1 percent of the score have roughly three times the average risk. These are solid population findings. What remains unproven is whether telling individuals their score improves health outcomes; the few small randomized trials have shown modest effects on cholesterol treatment uptake and little else.

Weak evidence (expert opinion, small studies, or marketing): consumer reports on nutrition, fitness, personality, wine preference, and most “wellness” traits. Statistical associations exist for some, but effect sizes are tiny and clinical validation is largely absent.

The professional consensus reflects this gradient. The American College of Medical Genetics and Genomics has stated that polygenic scores are not yet ready for routine clinical use, and the US Preventive Services Task Force limits its genetic recommendations to targeted testing for people with a qualifying family history. When a product’s language outruns this hierarchy, the language is the problem, not the evidence.

Whole genome sequencing vs exome vs array vs polygenic score: a comparison

The table below summarizes what each approach reads, what it is best at, and where it falls short. Keep in mind that a polygenic risk score is not a test technology at all; it is an analysis that can be run on data from any of the first three.

Approach	What it reads	Best for	Main blind spots	Evidence base
Whole genome sequencing	About 3 billion bases, coding and non-coding	Rare disease diagnosis; actionable monogenic variants; structural variants	Repeat expansions, pseudogene regions, epigenetics, interpretation gaps	Strong for diagnosis in symptomatic patients; limited for screening healthy adults
Whole exome sequencing	Protein-coding 1 to 2 percent of the genome	Most known disease-causing variants at lower data volume	Non-coding regulatory variants; uneven coverage from capture step	Strong for diagnosis; widely used clinically
Genotyping array (typical consumer kit)	Several hundred thousand pre-selected SNPs	Ancestry; common-variant risk; input for polygenic scores	Any rare variant not on the chip; false positives for rare variants	Reliable for common variants; poor for rare pathogenic variants
Polygenic risk score	Weighted sum of many common variants	Ranking inherited risk for common diseases across a population	Absolute risk, non-European ancestry, individual prediction	Moderate; outcome benefit not yet shown in randomized trials

One practical implication falls out of the array row. Research has found that a notable share of rare “pathogenic” calls from consumer arrays do not hold up when the same sample is sequenced in a clinical laboratory. Any array-based result pointing to a serious single-gene condition deserves confirmation by clinical sequencing before anyone acts on it.

Why a polygenic risk score can mislead, especially across ancestries

Two mathematical realities sit underneath every polygenic percentile, and consumer dashboards rarely spell them out.

The first is the gap between relative and absolute risk. Suppose a score places you in the top 10 percent for a disease that affects 1 percent of the population over a lifetime. Tripling that risk yields 3 percent. That is a real increase, and a clinician might reasonably factor it into screening conversations. It is also a 97 percent chance the disease never arrives. For coronary artery disease, which is far more common, the same percentile translates into a much larger absolute difference. The percentile alone cannot tell you which situation you are in.

The second reality is ancestry. Polygenic scores are built from association studies, and well over three quarters of participants in those studies have been of European descent. Because patterns of linked variants differ between populations, a score trained in one group loses accuracy in another. Analyses have shown that scores developed in European-ancestry cohorts perform only about a fifth to a half as well in people of African ancestry, with intermediate losses in East Asian, South Asian and Hispanic or Latino populations (NIH National Human Genome Research Institute). A score that is miscalibrated can place someone in a falsely reassuring or falsely alarming percentile.

Large efforts, including the NIH All of Us Research Program, are deliberately recruiting more diverse participants to narrow this gap, and newer multi-ancestry scores perform better than their predecessors. Until that work matures, the honest reading of any polygenic percentile includes a question: built from whom?

A third, quieter issue is that scores change. As studies grow, weights are updated and percentiles shift. A number from 2020 and a number from 2025 may disagree for the same person, not because the DNA changed but because the model did.

Will insurance cover whole genome sequencing, and who can see the result?

Coverage in the United States depends almost entirely on why the test is being done. When a clinician orders sequencing to diagnose a suspected genetic condition, or when a person meets established criteria such as a strong family history of early cancer or heart disease, many health plans cover the test, often after prior authorization. Consumer-initiated sequencing of a healthy adult, with no medical indication, is generally not a covered benefit (MedlinePlus Genetics). Policies vary by insurer, by state and by test, so the only reliable answer for an individual comes from the ordering clinic’s billing team and the plan itself.

The question underneath the coverage question is usually about discrimination. The federal Genetic Information Nondiscrimination Act of 2008, known as GINA, prohibits health insurers from using genetic information to set eligibility or premiums and prohibits employers with 15 or more workers from using it in hiring, firing or promotion decisions (MedlinePlus Genetics). The protection has gaps worth knowing. GINA does not apply to life insurance, disability insurance or long-term care insurance, and some states have added their own rules to fill those spaces while others have not.

Then there is data custody. A clinical result lives in a medical record under health privacy law. A consumer result lives on a company server under that company’s privacy policy and the laws of its state, which is a thinner shield. Most consumer platforms allow you to download your raw data and request deletion; the 2025 bankruptcy episode showed how many people had never looked at those settings. Reading a privacy policy before spitting in the tube is not paranoia. It is the same diligence you would apply to any record you cannot change and cannot take back.

Common myths about whole genome sequencing and polygenic risk scores

Viral claims about DNA testing tend to overshoot in both directions, promising certainty or dismissing the field entirely. Neither extreme matches the evidence.

Myth: a complete genome reveals every disease you will get. Sequencing reads inherited DNA. Most common diseases arise from an interplay of genes, environment, behavior and chance, and the genetic share is spread across thousands of small-effect variants. Twin studies, which estimate heritability by comparing identical and fraternal twins, put the inherited contribution to many common conditions at roughly 30 to 60 percent. The rest is not in the file.

Myth: a high polygenic score means the disease is coming. It means higher inherited risk relative to a reference group. Most people in the top decile of a polygenic score for coronary artery disease will not have a heart attack, and most heart attacks occur in people outside the top decile, simply because the middle of the distribution contains far more people.

Myth: a low score means you can skip screening. Mammography, colonoscopy, blood pressure and cholesterol checks are recommended by age and conventional risk factors. No guideline permits a consumer genetic score to override them.

Myth: consumer kits and clinical sequencing find the same things. A genotyping array reads a tiny, fixed fraction of the genome and has a documented false-positive problem for rare variants. The two are not interchangeable.

Myth: genetic data can be made fully anonymous. DNA is inherently identifying, and relatives share it. Studies have shown that a sizable share of Americans of European descent can be identified through distant relatives in public genealogy databases, whether or not they ever took a test themselves.

Myth: once the sequence is done, the answers are final. Variant classifications change as evidence accumulates. Reputable laboratories issue reclassification notices, which is one more reason clinical results, with a clinician attached, age better than app results.

When to see a doctor about genetic risk or a genomic result

Genomic information is only useful when it connects to a clinical conversation. The following situations are ones where that conversation should happen sooner rather than later, with a primary care clinician, a medical geneticist or a certified genetic counselor.

• A consumer or clinical report flags a pathogenic or likely pathogenic variant in a gene linked to cancer, heart disease or another serious condition. Treat an array-based finding as unconfirmed until a clinical laboratory verifies it.
• Your family history includes cancer diagnosed before age 50, several relatives with the same cancer, ovarian or male breast cancer at any age, sudden unexplained death in a relative under 50, or very high cholesterol in multiple family members. These patterns can qualify for clinician-ordered testing regardless of any consumer result.
• You are planning a pregnancy and carrier screening or a known family condition is relevant.
• A report lists a variant of uncertain significance and you are unsure what it means. The usual answer is reassurance plus a note in the record, but that judgment belongs to a clinician.
• A pharmacogenomic result suggests a medicine you take may be metabolized differently. Do not stop or adjust any prescribed medicine on the basis of a genetic report; bring it to the prescriber.
• A polygenic percentile is causing persistent anxiety or has led you to consider skipping recommended screening.

Red-flag symptoms that need prompt medical attention regardless of genetics include chest pain, fainting during exertion, a breast or testicular lump, unexplained weight loss, blood in stool or urine, and a first seizure. A genome result never changes the urgency of symptoms like these.

Every decision about screening intervals, preventive treatment or further testing should be made with the treating clinician, who can weigh the genetic finding against age, history and everything else that a sequence cannot see.

Frequently asked questions

What is the process of whole genome sequencing?

DNA is extracted from saliva or blood, sheared into fragments, read many times on a sequencing instrument, aligned against the human reference genome and then interpreted. Clinical labs usually read each position about 30 times to separate real variants from errors. A person differs from the reference at 4 to 5 million positions; interpretation narrows that to the handful with medical meaning, which is the slowest step.

What are the two methodologies of the Human Genome Project?

The public project used hierarchical shotgun sequencing, which mapped large DNA chunks to chromosomes before sequencing small pieces of each. A competing private effort used whole-genome shotgun sequencing, fragmenting the entire genome at once and relying on computing to reassemble overlaps. Both produced draft sequences announced in June 2000, and the whole-genome shotgun approach underlies every modern sequencer.

Will insurance cover whole genome sequencing?

Often yes when a clinician orders it for a medical reason such as a suspected genetic condition or a qualifying family history, usually with prior authorization. Consumer-initiated sequencing of a healthy adult is generally not covered. Policies differ by insurer and state, so confirm with the ordering clinic and your plan before testing.

What can whole genome sequencing not detect?

It commonly misses repeat-expansion conditions such as Huntington disease, variants in genes with near-identical pseudogene copies, low-level mosaicism, and all epigenetic changes like DNA methylation. It also cannot interpret variants of uncertain significance or account for environment, behavior and chance, which together explain most of the risk for common diseases.

What does a polygenic risk score mean for me personally?

It ranks your inherited risk for a common disease relative to a reference population, usually as a percentile. It is a population-level estimate, not a prediction for one person. Most people in the highest-risk group never develop the disease, and most cases occur in people with average scores. Discuss any score with a clinician alongside conventional risk factors.

Is whole genome sequencing better than exome sequencing?

It is more complete, not automatically better. Exome sequencing reads the protein-coding 1 to 2 percent of the genome where most known disease variants sit, at lower data volume. Whole genome sequencing adds non-coding regions and more even coverage of structural changes. For many diagnostic questions, both approaches reach the same answer, and clinicians choose based on the clinical situation.

Can a consumer DNA kit tell me if I carry a BRCA variant?

Only partially. Most consumer arrays check a small number of specific BRCA variants and can miss thousands of others. Studies have also found a meaningful false-positive rate for rare variants in array data. A positive consumer result should be confirmed by clinical sequencing, and a negative consumer result does not rule out hereditary cancer risk, especially with a strong family history.

How accurate are polygenic risk scores across different ancestries?

Less accurate outside European ancestry, because most association studies have enrolled European-descent participants. Analyses show scores developed in European cohorts perform roughly a fifth to a half as well in people of African ancestry, with intermediate losses in other groups. Multi-ancestry scores and more diverse research programs are improving this, but the gap has not closed.

Does the Genetic Information Nondiscrimination Act protect my genetic data?

GINA, a US federal law from 2008, bars health insurers from using genetic information to set eligibility or premiums and bars employers with 15 or more workers from using it in employment decisions. It does not cover life, disability or long-term care insurance, and it does not govern how consumer companies store or share data, which falls under their privacy policies and state law.

Should a healthy adult get whole genome sequencing for prevention?

No major guideline currently recommends it as routine screening for healthy adults, and the US Preventive Services Task Force limits genetic recommendations to targeted testing for people with qualifying family histories. Sequencing can identify actionable variants in a few percent of people, but whether population-wide sequencing improves outcomes is still being studied. The decision belongs in a conversation with a clinician or genetic counselor.

References

• MedlinePlus Genetics: What are whole exome sequencing and whole genome sequencing?
• MedlinePlus Genetics: Will health insurance cover the costs of genetic testing?
• MedlinePlus Genetics: What is genetic discrimination?
• NHS: Genetic and genomic testing

This article is for general information only and is not a substitute for professional medical advice. Please consult a qualified doctor about your individual situation.