The Polygenic Prophet: Hype and Reality of Polygenic Risk Scores

Polygenic risk scores (PRS) are the latest oracle in healthcare’s quest to predict fate. The basic idea sounds beguilingly simple: take the tiny genetic influences scattered across the genome and sum them up into one risk number. If one mutation can’t foretell disease, perhaps hundreds of thousands of them, weighed appropriately, can. In principle, a PRS condenses the polygenic nature of common diseases – the myriad of genetic variants with minuscule individual effects – into a single predictive index. The promise is a genomic crystal ball that might guide who gets cancer screening at 40 instead of 50, or who should watch their cholesterol like a hawk.

The concept has enchanted researchers and investors alike, spurring a decade of studies and startups. But as with any fashionable metric, a dry-eyed analysis reveals both genuine potential and ample room for skepticism. Is the PRS a revolutionary tool for personalized medicine, or just another overly quantified proxy – a high-tech risk horoscope of limited practical use?

What Exactly Is a Polygenic Risk Score?

In technical terms, a polygenic risk score is an aggregate measure of genetic liability to a trait or disease, computed by summing up the effects of many genetic variants across the genome (frontiersin.org). Each variant (usually single-nucleotide polymorphisms, SNPs) contributes a small weight to the score based on its association with the disease. Add them all up – often weighted by effect sizes from a genome-wide association study (GWAS) – and you get an individual’s polygenic score. Conceptually, if human traits are influenced by “many little things” in DNA, the PRS is an attempt to capture those many little influences in one number.

Mathematically, it looks like this:

$$
\text{PRS} = \beta_1 G_1 + \beta_2 G_2 + \cdots + \beta_N G_N
$$

G is the genotype (0/1/2 copies of a risk allele) at variant i,

and β is the effect size (log-odds or weight) of that variant on disease risk.

is the effect size (log-odds or weight) of that variant on disease risk . The effect sizes come from GWAS – those massive studies that scan genomes of thousands of people to find SNPs statistically associated with a disease. In plain English, a PRS is the summed contribution of many genetic dice-rolls, each only slightly loaded toward bad outcomes. The resulting score for a given person might be compared to a population distribution to see if they are in, say, the top 10% of genetic risk for a condition.

Importantly, because so many variants contribute, PRS values in a population tend to follow a Gaussian “bell-curve” distribution (frontiersin.org). Most people’s scores cluster around average, with relatively few at the high or low extremes. Having a high PRS for a disease doesn’t guarantee one will get sick; it simply means one’s genetic deck is stacked a bit more unfavorably than most. Similarly, a low PRS is no ironclad protection – it’s a statistical tendency, not fate.

As the American College of Medical Genetics (ACMG) dryly noted in 2023, “PRS test results do not provide a diagnosis, instead they provide a statistical prediction of increased risk” (geneticsandsociety.org). In other words, a polygenic score is more of a nudge from Dame Probability than a decree from the genetic fates.

To illustrate, consider a PRS for breast cancer. Researchers can take hundreds of common SNPs – none of which individually would worry even the most neurotic geneticist – and combine them into a score. The distribution of such a PRS in women, even those who go on to develop breast cancer, overlaps heavily with that of women who never develop the disease. The separation between “cases” and “controls” is a subtle shift of the curve rather than a dramatic split (frontiersin.org).

Figure 1 shows a typical result: the PRS distribution for breast cancer patients is slightly shifted to the right (higher) compared to controls, but with a lot of overlap (frontiersin.org). Genetics, it turns out, speaks in soft probabilities, not certainties.

Figure 1: Distribution of a breast cancer polygenic risk score in a population, comparing women who later developed breast cancer (red) vs those who did not (black). The curves overlap substantially, underscoring that PRS is a risk probability tool, not a deterministic test.

To avoid confusion, note that PRS goes by several aliases. It’s sometimes called a polygenic score (PGS) or genetic risk score (GRS) in the literature. The concept is the same. All these terms convey the essence that we’re aggregating many genetic signals into one risk metric. If a monogenic test (like checking BRCA1 for breast cancer risk) is a rifle shot, the PRS is a blunderbuss – wide-ranging and probabilistic.

One might wryly say that PRS is the triumph of quantity over quality in genetics: no single gene gives predictive power, so we conscript hundreds of thousands of them into service and hope their combined weight makes up in sum what each lacks in substance.

From GWAS to the Genome’s Crystal Ball: Academic Origins

To understand where PRSs came from, it helps to recall the arc of human genetics over the past 15–20 years. In the early 2000s, optimism was high that common diseases might have a few easy-to-find genetic triggers – the so-called “common disease, common variant” hypothesis. Researchers embarked on genome-wide association studies (GWAS) to sift genomes for hits, armed with new SNP microarrays that could genotype hundreds of thousands of variants at once.

By the late 2000s, GWAS had indeed uncovered genetic loci for many diseases (albeit often with dauntingly tiny effect sizes). Tens of thousands of associations have since been discovered for everything from diabetes to schizophrenia (nature.com). The upshot, however, was humbling: for most common ailments, no single SNP was destiny. Instead, dozens or hundreds of loci each conveyed incremental risk changes (odds ratios often in the 1.1–1.3 range – barely a whisper above noise).

It was in this context that the concept of polygenic scoring emerged. If lots of variants each have minuscule impact, perhaps one can aggregate them to get a stronger signal. Early pioneers like the International Schizophrenia Consortium (2009) showed that even SNPs not reaching formal GWAS significance could collectively differentiate cases and controls – evidence that disease risk is highly polygenic and that scores summing many sub-significant SNPs carried predictive information (nature.com). Naomi Wray and others formalized methodologies for constructing and validating these scores in the early 2010s, giving the approach a statistical backbone.

As data grew, so did PRS performance. The launch of large biobanks (UK Biobank with ~500,000 individuals of genomically somewhat homogeneous British descent, among others) provided massive discovery and validation datasets. Studies began to report that a PRS could sometimes identify subsets of the population with substantially elevated risk. A watershed moment came in 2018, when a team led by Amit Khera showed that polygenic risk scores for several diseases (coronary artery disease, diabetes, etc.) could single out the “genetic 8%” – people with risk equivalent to that conferred by a rare monogenic mutation (nature.com).

For example, their optimized PRS for coronary artery disease (based on millions of variants across the genome) tagged about 8% of individuals as having more than triple the normal risk of a heart attack – a risk on par with carrying a pathogenic mutation in a single gene like LDLR or PCSK9 (nature.com). This was a splashy finding: it suggested that while no one SNP predicts heart attacks in the general population, the combined effect of many SNPs could – in theory – identify as many high-risk people as traditional monogenic testing (albeit different people, with more modest per-person risk). “Suddenly, polygenic scores could find ‘genetic high risk’ individuals that were invisible to classical medical genetics,” the narrative went. Calls to integrate such scores into clinical care soon followed.

Meanwhile, the first decade of PRS research (roughly 2008–2017) taught important lessons. One was about ancestry and applicability. Most early GWAS and PRS studies focused on individuals of European ancestry, a matter of convenience and cohort availability but one with serious downstream implications. An analysis in Nature Communications found that 67% of polygenic score studies up to 2017 were done exclusively in Europeans, 19% in East Asians, and under 4% in populations of African, Hispanic, or Indigenous descent.

This gross under-representation meant that PRS models were being built and tuned on a very narrow slice of human diversity. As we’ll explore, a Eurocentric genie in the bottle may not grant wishes to those of non-European genetic backgrounds; indeed, applying a PRS trained in one ancestry to another can drastically erode its predictive power. The academic community gradually woke up to this limitation and, in recent years, has prioritized more diverse GWAS and cross-population methods (nature.com). But the bias in the foundational data is still being reckoned with.

Another piece of the origin story involves technological and statistical advances. Beyond simple “clump and threshold” methods of score-building (take all SNPs below some p-value, weight by GWAS beta, ignore those in high linkage disequilibrium, etc.), researchers developed increasingly sophisticated algorithms – Bayesian shrinkage methods, LDpred, Lassosum, PRSice, and so on. These improved how one selects and weights SNPs for a score, aiming to squeeze out a bit more signal-to-noise. For the true connoisseurs, the field became a contest of polygenic prognosticators: whose PRS method could achieve the highest R² or area under the curve (AUC) in an independent sample?

By the time the UK Biobank and other large datasets became available, PRS construction turned almost plug-and-play. Online catalogs now host pre-made scores for dozens of diseases (e.g. the Polygenic Score Catalogfrontiersin.org), ready for researchers to apply to their cohorts. What once required a team of statistical geneticists can now be done by an enterprising graduate student with the right software and a big spreadsheet of SNP genotypes.

In summary, the PRS concept emerged as a logical extension of GWAS: having failed to find simple genetic causes for complex diseases, scientists shifted to aggregating weak signals into a single index of genetic risk. It’s a bit like admitting no single straw will break the camel’s back, so you gather a thousand straws and hope their combined weight might at least predict which camels are in trouble.

The academic foundation was laid by improvements in genotyping, massive cohorts (biobanks), and increasingly clever statistical tricks to maximize predictive mileage from mountains of SNP data. By the late 2010s, polygenic risk scoring had moved from fringe idea to genomic zeitgeist, crossing over from research journals to the business plans of healthcare startups.

Present Applications: How PRS Is (and Isn’t) Used Today

What can one actually do with a polygenic risk score in 2025? At present, PRSs straddle an awkward line between research tool and clinical experiment. Enthusiasts see them as heralding the era of predictive, personalized medicine; skeptics note that routine medical practice hasn’t exactly been revolutionized yet. Let’s survey a few domains where PRS has made inroads – cancer risk, neurological and psychiatric disorders, and some emerging use-cases – to see how the score is being deployed.

Cancer Risk Stratification: From Breast to Colorectal

On the cancer front, breast cancer is often cited as the poster child for PRS utility. This makes sense: breast cancer has well-known high-risk genes (BRCA1/2), but those are rare; most cases arise in women with no monogenic mutation. A polygenic score, which captures the combined impact of many common variants, offers a way to stratify risk in the vast majority of women who are “genetically unremarkable” by traditional testing. Over the past decade, researchers have built increasingly powerful PRSs for breast cancer.

For example, one widely studied model uses 313 SNPs and has been validated in large prospective cohorts (breast-cancer-research.biomedcentral.com). This PRS313 yields about a 1.6-fold increase in breast cancer risk per standard deviation of the score. In practical terms, women in the top few percentiles of the score have several-fold higher risk than women in the bottom few percentiles. An illustrative statistic: Women in the highest 1% of the PRS distribution had an estimated 32.6% lifetime risk of developing breast cancer, compared to ~2% for those in the lowest 1%. That’s a striking gradient – roughly a 16-fold difference in absolute risk between extremes – albeit spanning from “very low” to “moderately high” risk in absolute terms (for context, average female lifetime risk of breast cancer is about 12%).

Such stratification opens the door to personalized screening strategies. If a woman’s PRS puts her in, say, the top 5% of risk (perhaps equivalent to the risk of someone 10–15 years older with average genetics), one could argue she should start mammograms earlier or consider preventive measures like enhanced MRI screening or even chemoprevention. In fact, clinical decision tools are being updated to incorporate PRS. The CanRisk model (an update of the IBIS/Tyrer-Cuzick model) now allows input of a polygenic score alongside family history and other factors. Myriad Genetics, a company known for hereditary cancer testing, offers an integrated test called RiskScore that combines an 86-variant PRS with traditional risk factors to provide an individualized breast cancer risk for women who don’t have a BRCA mutation (investor.myriad.com).

In their studies, this combined approach reclassified a substantial subset of women into higher or lower risk categories, refining who might qualify for intensified surveillance (cancernetwork.com). Notably, Myriad’s RiskScore had to confront the ancestry issue – it was initially calibrated in European women, but the company developed separate PRS models or adjusted algorithms for African, Asian, and Hispanic ancestries to make it valid “for all ancestries” (a claim that underwent scrutiny, since non-European performance still lags somewhat). Nonetheless, the first tentative steps toward clinical use of PRS in breast cancer are underway, with international expert groups publishing guidelines on how to incorporate PRS into risk prediction and screening decisions (mdpi.comgeneticsandsociety.org).

Beyond breast cancer, prostate cancer has seen PRS-based pilot programs. One innovative trial in the U.K. (the BARCODE1 study) took a population screening approach using PRS. Men in their mid-50s were invited to give a spit sample for genotyping; those with a PRS in the top 10% for prostate cancer risk were offered an MRI and biopsy regardless of their PSA level (ascopost.comascopost.com).

The rationale was that genetics might identify high-risk men who would be missed by PSA screening. The results were intriguing: among the men with top-decile PRS who underwent proactive screening, 40% were found to have prostate cancer (many with intermediate or high-risk tumors that warranted treatment).

Over 70% of those significant cancers would not have been flagged by the usual PSA-plus-MRI pathway. In other words, the PRS-led approach fished out tumors that the conventional screening net was missing in those younger men. The study concluded that targeting the top 10% genetic risk achieved higher yield of significant cancers than standard practice.

It’s a proof-of-concept that PRS could help reshape screening paradigms – concentrating expensive tests (like MRIs and biopsies) on those genetically most likely to benefit. Of course, this was a research trial; we are not yet at the point where every 55-year-old man gets his polygenic score checked at the GP. But it points to a possible future where an EHR might flash a “high genetic risk” alert that influences a screening plan.

Colorectal cancer is another case in point. Like breast, general guidelines use age as the blunt criterion for everyone. But researchers in Finland and elsewhere have shown that a PRS for colorectal cancer can stratify risk sufficiently to consider adjusting colonoscopy schedules.

One recent analysis suggested that, between someone in the highest PRS decile and someone in the lowest decile, a sensible risk-tailored difference in screening start age could be on the order of two decades (pmc.ncbi.nlm.nih.gov). For instance, men with top 10% polygenic risk might warrant screening in their 40s, whereas women with bottom-tier risk could safely wait until their 60s.

Another study presented at a European genetics conference found that in Finland, using PRS could imply as much as a 16-year difference in when to start colonoscopies between high-risk vs low-risk individuals (ecancer.org). These are still theoretical models – no health system has yet overhauled its screening program to be gene-based – but they illustrate the promise. Public health folks are understandably cautious: implementing such changes would require ironclad evidence from trials showing improved outcomes and cost-effectiveness.

Those studies are likely forthcoming. In the meantime, it’s telling that the European Society of Human Genetics and national research programs are actively discussing PRS-guided screening for cancers.

Finally, it’s worth noting an intriguing use of PRS in gene carriers. Even when someone has a high-risk mutation like BRCA1, not everyone with the mutation has equal risk – there is variability, partly due to genetic background. Polygenic scores can potentially refine risks for such individuals. For example, among women who carry a BRCA1 mutation, those who also happen to land in the top decile of a breast cancer PRS have been reported to have double-digit higher absolute risks of developing cancer by age 50 than those in the bottom decile (nature.com).

One study found BRCA carriers in the highest PRS decile had a 39% risk by age 50, compared to 21% if they were in the lowest decile (nature.com). This kind of information could inform decisions about preventive mastectomy or intensified surveillance – effectively using PRS to modulate management even in “genetically heavy” cases. It’s an example of how polygenic background can modify penetrance of monogenic risk, a topic of growing research interest.

Stepping back, where do we stand in cancer? PRSs for breast, prostate, and colorectal cancer (and others like melanoma or ovarian to lesser extents) are poised at the brink of clinical utility. They are being integrated into risk models, tried out in screening studies, and contemplated by guideline-makers. But broad clinical adoption is still cautious and limited. One reason is that clinical thresholds are tricky: What do you do for someone whose

PRS puts them at 2× the normal risk? (If normal risk is small, 2× is still small.) There’s also the matter of intervention – identifying high risk is only useful if there’s something effective to offer. In breast and colorectal cancer, there are clear screening/preventive options, which is why those are forefront. In other cancers, say pancreatic or brain cancer, knowing one’s high polygenic risk might just induce dread since early detection tools are lacking. Thus, PRS deployment is currently targeted to scenarios where actionable steps exist.

Brain and Behavioral Conditions: Schizophrenia, Alzheimer’s, and Beyond

Polygenic risk scores found some of their earliest successes in the realm of psychiatric and neurological disorders. The first published PRS demonstration was in schizophrenia: the International Schizophrenia Consortium in 2009 showed that common SNPs collectively could distinguish cases from controls better than chance, essentially proving schizophrenia has a polygenic architecture (nature.com).

Since then, PRSs have been developed for most major brain-related conditions, from depression to bipolar disorder to autism. In psychiatry, PRS has become a standard research tool – a way to quantify genetic liability and see how it correlates with clinical or biological markers. For instance, a higher schizophrenia PRS correlates not just with schizophrenia itself, but also with certain cognitive or neuroimaging traits even in healthy people (suggesting the score captures some latent aspects of brain function).

Clinically though, the utility is less clear. Schizophrenia PRS can identify people with elevated risk, but no consensus yet exists to screen the general population for psychiatric risk – and what would one do if a 18-year-old had a top 0.1% PRS for schizophrenia? There’s no proven preventive intervention (we can’t exactly start prophylactic antipsychotic drugs).

Still, one could envision PRS being used to enrich clinical trials or identify high-risk individuals for closer monitoring (e.g. young people with subtle symptoms plus high PRS might benefit from early therapy). As of now, these ideas remain experimental. Some have half-joked that using a PRS for a psychiatric illness in clinic is like a horoscope: it might spur conversation, but you wouldn’t want to make life decisions on it alone.

In Alzheimer’s disease, the situation is a bit different because there are known strong genetic factors (namely APOE ε4 allele) and also an emerging landscape of preventive trials. A PRS for Alzheimer’s typically includes APOE plus many other SNPs. Such a score can stratify risk for cognitive decline or AD conversion among older adults. For example, individuals with high AD polygenic risk (and especially if APOE4-positive) have much higher odds of developing Alzheimer’s by a given age than those with low risk scores. Some observational studies and trials have started incorporating PRS to select participants.

A high-profile example: the Generation Program trials a few years ago targeted people who were APOE4 carriers (high genetic risk) to test an Alzheimer’s prevention drug – they didn’t include PRS per se, but one can easily imagine adding a genome-wide PRS to refine eligibility. Researchers have found that PRS can differentiate Alzheimer’s cases from controls in biobank data, and even correlate with progression or clinical heterogeneity (nature.com).

However, analogous to psychiatry, the clinical use of an Alzheimer’s PRS for an individual patient is debatable. Would a 50-year-old want to know they have a high polygenic risk of Alzheimer’s when currently there’s no surefire way to stop the disease? (Some might, for planning or motivation to modify lifestyle; others very much would prefer not to know.)

With new monoclonal antibody treatments emerging that can modestly slow early Alzheimer’s, one could foresee PRS being used in the future to identify at-risk cognitively normal people for early intervention – but we are not quite there yet. Direct-to-consumer testing companies do sometimes give customers reports like “increased genetic risk for late-onset Alzheimer’s” based on APOE and PRS, with heavy disclaimers. It’s an area that veers quickly into ethical grey zones about anxiety vs empowerment.

Interestingly, one of the research uses of PRS in neurology/psychiatry has been to probe the genetic overlap between disorders. For example, the PRS for schizophrenia is also modestly associated with bipolar disorder and even major depression, indicating a shared genetic underpinning. PRS for one condition can often predict another at above-chance levels – a phenomenon known as pleiotropy.

This has yielded insights like: autism and schizophrenia PRSs are largely uncorrelated (genetic risk that predisposes to one does not predispose to the other, despite both being developmental brain disorders). Or that PRS can help subclassify disorders: an epilepsy PRS might differ between subtypes of epilepsy (nature.com). These are fascinating from a science perspective, though not yet directly altering patient care.

In summary, for CNS diseases and psychiatric conditions, PRS is a powerful research instrument and holds future potential for risk stratification. But today, it’s not standard of care to genotype every person for these scores – unlike cancer or heart disease, there’s typically no immediate actionability. Telling someone they have a high genetic risk of schizophrenia is a bit like telling them they have a stormy fortune in tea leaves – it might be true, but what are they to do with that information?

This doesn’t mean it will never be useful; it just means we need either preventive measures or early interventions specific to those risks in order to make it worthwhile. There is ongoing work, for instance, in using PRS to predict treatment response (e.g. maybe people with high polygenic risk for depression respond differently to certain antidepressants – preliminary evidence is scant, but it’s being looked at). So the clinical utility is on the horizon, glimmering but not fully formed.

Other Emerging Uses: Heart Disease, Diabetes, and More

Though the question focuses on cancer and CNS, it’s worth briefly noting that cardiovascular disease and metabolic disorders are another active area of PRS application. Indeed, coronary artery disease (CAD) was one of the earliest touted examples where a PRS could pick out high-risk individuals (as noted with Khera’s 2018 study). The idea of integrating PRS into cardiac risk scoring is being vigorously explored. Unlike in psychiatric disease, here there are preventive actions to take (statins, blood pressure control, etc.), so the question boils down to: does knowing someone’s genetic risk improve on conventional risk factors enough to change management?

So far, studies have found that a CAD PRS is an independent risk factor – if you add it to traditional clinical models (age, cholesterol, blood pressure, etc.), it statistically improves predictions, but usually by a modest amount (pmc.ncbi.nlm.nih.gov). For example, one large UK Biobank analysis showed adding a polygenic score to the standard risk calculator (Pooled Cohort Equations) improved the area-under-curve from about 0.755 to 0.776 – a small bump (pmc.ncbi.nlm.nih.gov).

The net reclassification of patients into correct risk categories was also small but significant (a few percent of people were re-stratified appropriately). Critics note that a 1–2% improvement in predictive accuracy is hardly a game-changer in practice. On the other hand, proponents argue that for the subset of people at extremes of genetic risk, the information is very relevant. Someone in the top 1% PRS for heart disease has an odds ratio ~3–5× the average risknature.com, which might justify earlier or more aggressive preventive therapy. Conversely, a young person with very low polygenic risk might not need interventions as urgently if other factors are borderline.

Clinically, no guidelines yet say “use a PRS to decide on statin therapy,” but we are inching in that direction. The American Heart Association issued a 2022 scientific statement reviewing PRS for cardiovascular disease, essentially concluding that it’s promising but not fully validated for routine care (ahajournals.org). Some trials have tested the psychological and behavioral impact of giving people their PRS for CAD – one called MI-GENES randomly gave half of participants a genetic risk score in addition to traditional counseling to see if it would motivate them to change lifestyle or adhere to statins.

The result? There was a slight improvement in LDL cholesterol in those informed of their genetic risk, hinting that PRS might nudge some patients toward better compliance or risk factor control. But the effect was not dramatic, and 10-year follow-up suggested no major difference in outcomes (ahajournals.org).

In diabetes, 23andMe made headlines by launching a direct-to-consumer Type 2 Diabetes Polygenic Risk report in 2019 (statnews.com). It uses a PRS of over 1,000 variants and even incorporates self-reported BMI into the algorithm (a slightly cheeky move – mixing genetics with environment to boost predictive power – which purists criticized). The idea was to alert individuals if they have a significantly elevated genetic risk for diabetes, with the hope they might adjust their diet or weight accordingly.

However, the utility of this is questionable, as weight and family history are already obvious risk indicators. Indeed, one might ironically note: most people at high genetic risk for type 2 diabetes learn something they already knew if they look in the mirror or at their family tree. The genetic score correlates with family history and population/ethnic background to a degree. Still, some evidence suggests polygenic scores could identify a subset of normal-weight people who are at hidden higher risk for diabetes, or separate out truly low-risk obese individuals from high-risk obese individuals (pmc.ncbi.nlm.nih.gov). Research is ongoing to see if genetic info can refine diabetes prevention targeting.

Even outside the realm of health and into traits, there’s buzz (and controversy) around PRS. Scores have been built for things like educational attainment or intelligence, height, body mass index, etc. These have no clinical application per se but represent the power and peril of polygenic prediction. The accuracy for something like adult height is actually fairly high (R² ~40% for Europeans), showing that if a trait is sufficiently heritable and measured well, a PRS can capture a lot of that genetic variance. That’s a scientific achievement, but also a bit of a party trick – nobody needs a DNA test to tell how tall their child will likely be (there’s a low-tech way: measure the parents!). Similarly, a PRS for IQ or educational level raises alarms about genetic determinism and misuse, which leads us to the next section on limitations and ethical concerns.

The Promises of PRS: Early Screening, Personalized Medicine, and More

Before we delve into the pitfalls, let’s acknowledge why PRS has garnered so much excitement. Proponents of polygenic scores paint a vision of more proactive, personalized healthcare, enabled by genetic insights. Here are some of the oft-touted promises of PRS:

Earlier and More Targeted Screening

Perhaps the clearest near-term benefit. If you know someone is in the top few percent of genetic risk for a cancer or heart disease, you could begin screening or preventive measures sooner. This might catch disease in early stages or even avert it. We saw examples in cancer where PRS could shift start ages for mammograms or colonoscopies. In cardiovascular disease, a high-risk person might get their coronary calcium scan earlier or more frequent monitoring of risk factors.

Essentially, PRS offers a way to stratify populations into risk tiers beyond what traditional factors provide (nature.com, genomicseducation.hee.nhs.uk). Rather than one-size-fits-all at age 50 or 60, genetics could help define who among, say, 40-year-olds, already has the risk of a typical 60-year-old. Catching those folks could enable “pre-emptive strikes” – intensified screening or prophylactic medications – while sparing truly low-risk folks from unnecessary procedures. In public health terms, polygenic scores might improve the efficiency of screening programs, focusing resources on those most likely to benefit (ecancer.orgecancer.org).

Personalized Preventive Medicine

Hand in hand with earlier screening comes the idea of tailoring prevention. For example, if a patient has a high PRS for coronary disease, a doctor might be more aggressive about starting a statin or blood pressure drug even if the patient’s current levels are moderate. Conversely, a low genetic risk person might manage with lifestyle changes longer before medications.

The vision is that PRS becomes one more factor (like blood pressure or cholesterol) in the risk-profile equation, enabling finer personalization of care (frontiersin.org, genomicseducation.hee.nhs.uk). In oncology, a woman with high polygenic risk might opt for preventative tamoxifen or even prophylactic surgery despite no single high-risk gene mutation – something currently done mainly for BRCA carriers. In essence, PRS could identify people who have “unknown” risk that would qualify them for existing preventive interventions.

Notably, the interventions themselves aren’t new – it’s the targeting that would improve. We already have cholesterol-lowering drugs, diet/exercise advice, chemoprevention, etc.; PRS would refine who gets an intervention, ideally improving the benefit-risk tradeoff and cost-effectiveness.

Cohort Enrichment for Trials and Studies

This is a less publicized but very practical promise. Clinical trials for preventive therapies or screening methods often need to enroll a lot of people to observe enough disease events, especially if the disease is relatively rare or takes a long time to develop. By selecting participants with higher genetic risk, trials can enrich the event rate, making it easier and faster to see if an intervention works. For example, a trial of a new Alzheimer’s-preventive drug might recruit people in their 60s who not only carry APOE4 but also have a high polygenic risk score – increasing the chance that a good proportion will develop cognitive decline in the follow-up period, thus testing the drug’s efficacy more efficiently.

Similarly, trials of intensive breast cancer screening could target those with high PRS (plus other risk factors), because they are the ones who might actually develop cancer at a young age and demonstrate whether intensified screening helps. This use of PRS doesn’t directly involve patients knowing their scores at all – it’s a research design tool to make studies cheaper or more powerful. It’s already happening in some contexts, such as using PRS as part of eligibility for clinical trials in cardiovascular prevention and Alzheimer’s (often in combination with family history or specific genes). The “cohort enrichment” promise might not grab headlines, but it’s one way PRS could indirectly accelerate the development of new therapies by concentrating research on higher-risk groups (nature.com).

Enhancing Risk Models and Precision Health Algorithms

In the broader sense, PRS is frequently touted as a key component of the evolving field of precision medicine. Think of the algorithms that might soon live in electronic health records: these will integrate various data – demographics, medical history, lifestyle, maybe wearable device data – to generate individualized risk scores for patients. Genetics (including PRS) is an important piece of that puzzle, adding information that is lifelong and often independent of environment.

Already, life insurers and researchers talk about “integrated risk tools” where, for example, a 45-year-old’s 10-year heart attack risk is calculated by combining their clinical risk score (like Framingham) and their polygenic risk score (genomicseducation.hee.nhs.uk). One company, Genomics plc, working with the UK’s NHS, demonstrated a prototype where such an integrated score reclassified some people’s risk enough to change management – catching high-genetic-risk individuals who would have been deemed low-risk by clinical factors alone.

The promise here is more precise risk stratification leading to more tailored recommendations in healthcare. It’s the holy grail of moving from population-level guidelines (“everyone over 50 do X”) to individual-level guidance (“you, given your unique risk profile, should start at 47, while your neighbor can wait till 60.”).

Understanding Biology and Identifying Novel Targets

Though not as prominently discussed as a “promise” for immediate health impact, an important scientific benefit of PRS is what it reveals about disease biology. By capturing the genetic component of a disease in a single variable, researchers can ask interesting questions: What else correlates with the PRS? For instance, does a high PRS for heart disease correlate with certain blood biomarkers? If so, those might be causal pathways worth targeting. PRS can also be used in Mendelian randomization studies (a method to test causal relationships using genetics as instrumental variables).

Because a PRS is essentially a genetic instrument for disease propensity, you can, for example, use PRS for BMI to test if higher lifetime BMI causally leads to worse outcomes – without the confounding typical of observational studies. Additionally, PRS can help identify subtypes of disease; if some patients have high PRS and others low PRS, do they respond differently to treatments or have different disease trajectories? In schizophrenia, for instance, those with very high PRS might represent a more strongly genetically loaded form of the illness, which could have implications for prognosis or treatment response (an area of ongoing research). While these aspects are more research-oriented, the long-term promise is that PRS-driven insights could point to new drug targets or more personalized therapeutic strategies, dividing diseases into genetic subgroups.

Patient Empowerment and Engagement

This one is a bit controversial. Some argue that giving people information about their genetic risk can empower them to take charge of their health – motivating them to change their lifestyle or adhere to medications. Skeptics worry it can also lead to fatalism or anxiety. The evidence so far is mixed. But in an optimistic framing, a person who learns they have a high PRS for type 2 diabetes and a family history might feel a renewed commitment to weight management, more so than if they were just told “you should lose weight because it’s healthy.”

The genetic component adds a sense of personalization – “I’m at risk, this is about me, not just general advice.” Studies have found moderate psychological impact: many people find it interesting, some get motivated, a few might overreact. It’s certainly not a panacea for behavior change. Still, in the arsenal of tools to improve patient engagement, PRS is another potential lever – ideally used with proper counseling to inspire positive action rather than dread.

These promises certainly have a rosy tint, and reality will likely temper them. But it’s fair to say that polygenic risk scores have opened new avenues in both research and preventive medicine that were not available before. We can now identify a larger fraction of individuals at high risk for common diseases (albeit with probabilistic predictions) than was possible when we only looked for single gene mutationsnature.com. If integrated wisely, this could mean shifting from reactive to proactive care for some conditions, and more efficient focusing of medical resources. In healthcare systems increasingly interested in prevention (to reduce long-term costs and morbidity), that is an attractive proposition.

The strategic framing of PRS by its proponents is that of a risk management revolution: not replacing traditional risk factors, but adding a new dimension that was previously hidden. We already manage risk factors like blood pressure or smoking; genetic risk is just another piece of the risk pie chart – one that you’re born with, so it’s there to measure long before disease manifests. Why not use it? – so goes the logic. Indeed, some industry analyses estimate that widespread use of PRS could prevent thousands of heart attacks or cancer cases via better targeting of prevention. It all sounds great on PowerPoint slides.

What’s the catch? If this is so wonderful, why aren’t doctors ordering polygenic scores for every patient already? To answer that, we must examine the many limitations, challenges, and outright risks of misuse that shadow the PRS excitement.

Limitations and Challenges of Polygenic Risk Scores

The glowing promises of PRS come with equally significant caveats. Like any tool, polygenic risk scoring has limitations – some intrinsic to the science, others related to ethical and practical issues. A savvy investor or clinician would do well to consider these before betting the farm on a future of genomic fortune-telling. Let’s dissect the key limitations:

Narrow Ancestral Applicability – The Eurocentric Bias

Perhaps the most glaring limitation is that PRSs often perform poorly outside the populations in which they were developed. The vast majority of GWAS (which feed into PRS construction) have been done in people of European descent As a result, a PRS that predicts reasonably well in, say, a British population, may have much lower accuracy in an African-American or South Asian individual (nature.com).

Empirically, studies have shown dramatic drops in predictive power: one analysis found that a European-derived PRS had only about 42% of its original effect size when applied to African-ancestry samples (nature.com). East Asian populations fared a bit better (around 95% in that analysis) and South Asians around 60%, but the pattern is clear – the more genetically distant the target population from the discovery population, the weaker the score performs.

The reasons are multifold: differences in allele frequencies, differences in linkage disequilibrium patterns (how SNPs correlate with each other), and the presence of population-specific risk variants that wouldn’t have been tagged in a Euro-centric GWAS. In plain terms, a score built on one ancestral group is somewhat miscalibrated in another – it might systematically under- or over-estimate risk. For example, a European PRS might give systematically lower values to African ancestry individuals due to different genetic architecture, falsely suggesting they’re at low risk when they are not (or vice versa).

This is a huge equity issue: using current PRS in the clinic could exacerbate health disparities by offering better prediction for some groups and misleading information for others (nature.com). The field is actively working on solutions – initiatives to collect GWAS data in diverse cohorts, statistical methods to adjust or fine-tune scores for different ancestries (nature.com) and multi-ancestry polygenic scores that combine data from multiple populations (prnewswire.com).

Progress is being made (e.g. there are now improved scores for CAD that include non-European data, boosting performance in South Asians and others (committees.parliament.uk).

But as of today, many off-the-shelf PRS are still Eurocentric. The ACMG and others have explicitly cautioned that applying PRS across ancestries can be unreliable (genomicseducation.hee.nhs.uk). It’s a bit of a catch-22: the people who might benefit most from better risk prediction – historically marginalized populations – are the ones for whom current PRSs are least useful. Until this gap is closed, PRS could inadvertently worsen disparities if used incautiously (for instance, if healthcare providers only validate a score in Europeans and then use it on everyone, non-European patients could get misleading risk info or be mis-triaged).

Unclear Clinical Utility & Actionability – When and How to Use a Score?

Even if a PRS is technically predictive, it doesn’t automatically follow that it’s clinically useful. The threshold question is: At what point does a predicted risk warrant a clinical action? For monogenic mutations, we have some benchmarks (e.g. if lifetime breast cancer risk > 20%, recommend MRI screening; if LDL cholesterol above a certain genetic threshold, treat as familial hypercholesterolemia, etc.). For PRS, those lines are fuzzy. If a patient is in the top 5% polygenic risk for heart disease, do we put them on a statin regardless of their LDL level? Some experts might lean yes, others are hesitant because clinical trials haven’t specifically confirmed that approach.

No major clinical guidelines yet include polygenic scores as a deciding factor for treatment. In part that’s because we lack long-term evidence that using PRS to guide interventions improves outcomes. The PHG Foundation report in 2019 bluntly concluded there was “no evidence for the clinical utility of PRS for cardiovascular disease prevention at that stage” (genomicseducation.hee.nhs.uk) – not saying it won’t have utility, but that it hadn’t been proven yet. Another facet is that many PRSs have modest predictive power: an AUC of 0.6–0.7 in population prediction, for instance.

That means while better than random, they are far from a crystal ball. For diseases that already have other risk models, the incremental value of PRS might be small. Clinicians today are juggling many risk scores (for heart disease, for cancer, etc.) based on clinical factors; adding a genetic score is only helpful if it changes management in a justifiable way. If it doesn’t change what you would do, then it’s at best unnecessary information and at worst confusing or misleading. Right now, outside of research or very specialized clinics, most doctors wouldn’t know what to do differently with a patient’s PRS result.

This is starting to change in some areas (as we noted, pilot guidelines for breast cancer screening with PRS are being drafted, and some cardiology consensus statements discuss what to “consider” if a PRS is known (ahajournals.orgahajournals.org).

But by and large, lack of consensus on how to integrate PRS into practice is a bottleneck. Until we have clear, evidence-based thresholds (e.g. “if 10-year risk ≥ X% by combined model including PRS, then do Y intervention”), many providers will be rightly cautious. There’s also the risk of over-testing or over-treating based on genetic risk that might never materialize into disease – the false positives problem.

If a score is not highly specific, we might flag a lot of people as “high risk” who ultimately would never fall ill, subjecting them to stress and possibly interventions that only benefit a few. Fine-tuning PRS utility thus requires more research and perhaps a reframing: some argue PRS should be reserved for situations where it clearly tips the scale on a management decision that could go either way (like borderline cases). Otherwise, it’s just an interesting number.

Genetic Determinism and Ethical Misuse

With great predictive power comes great responsibility – or something like that. One big concern around polygenic scores is how they might feed genetic determinism in the public mind or be misused in various ways. The notion that a single number can summarize your genetic destiny for a trait is powerful – and potentially misleading. People may misunderstand probability as fate. For instance, someone with a high Alzheimer’s PRS might think they are doomed and disengage from life, when in fact it’s far from certain they’ll develop the disease (especially with preventive measures). On the flip side, someone with a low cancer PRS might get overconfident and neglect screenings or healthy behaviors (“Why exercise? My DNA says I’m low risk for heart disease!” – a dangerous fallacy, since low genetic risk is not no risk).

The ACMG’s point #2 in their 2023 statement emphasizes this: a low PRS does not rule out significant risk from other factors(geneticsandsociety.org). Ethically, there’s fear that PRS could be used for nefarious purposes like genetic discrimination. Could insurers or employers one day demand your polygenic scores? In many jurisdictions, laws like GINA (Genetic Information Nondiscrimination Act) provide some protection for health insurance and employment – but not for life insurance or long-term care insurance, for example.

If actuaries believe a PRS for longevity or disease risk is valid, they might push to use it in underwriting. This raises questions of fairness: these scores are probabilistic and often correlate with ancestry, so using them could inadvertently become a proxy for discriminating against certain ethnic groups if not handled carefully. Another area of ethical contention is embryo selection.

Startups have already begun offering polygenic testing of IVF embryos – claiming they can select embryos with lower risk of adult diseases (or even higher potential IQ). This is highly controversial. Leading genetics bodies have forcefully argued that using PRS to pick embryos is unwarranted and unethical at this time (geneticsandsociety.org).

The ACMG explicitly stated that “preimplantation PRS testing is not yet appropriate for clinical use and should not be offered”. Why? Because the predictive power is low, the trade-offs are unclear, and it veers into eugenics territory. Imagine choosing one embryo over another because it has a 5% vs 4% projected risk of heart disease – a difference that could be rendered moot by lifestyle. Yet, fertility clinics and companies are marketing these services, with “wildly inappropriate marketing promises” according to ethicists (geneticsandsociety.org).

The specter of designer babies, or at least “curated” babies, based on polygenic traits has caused justifiable concern. Beyond embryos, even in adults, the idea of summing up a person by polygenic scores for various traits (intelligence, mental illness propensity, etc.) raises dystopian scenarios if misused by authoritarian governments or unscrupulous entities. This is where the dry, ironic perspective is healthy: one must remember that a PRS is a flimsy predictor, not an engraved destiny, and treating people as more than just their genome is crucial.

Direct-to-Consumer (DTC) Hype and Consumer Misunderstanding

Polygenic risk scores have already been commodified to an extent by DTC genetic testing companies. While some, like 23andMe, have been relatively careful (their FDA-approved health reports include a Type 2 Diabetes PRS, but with lots of education), others have been less so. There are third-party services where one can upload raw genetic data and get a smorgasbord of risk scores for dozens of diseases. The accuracy and validation of these are questionable, and consumers may not have the guidance to interpret results. The potential for misinterpretation is high. For example, if a report says “Your polygenic risk for coronary artery disease is in the 85th percentile (moderately elevated),” a consumer might panic or, conversely, dismiss it if they feel fine.

They might not realize that “85th percentile” might only shift absolute risk from, say, 5% to 8% – significant in epidemiological terms but not a doomsday verdict. On the other hand, someone at 20th percentile might falsely believe they’re “heart-attack-proof” and neglect sensible precautions. The communication of PRS results is a delicate art, and DTC channels often lack the counseling component.

The ACMG’s 2021 points to consider on DTC testing highlighted that complex trait testing is an “inexact science” and not equivalent to monogenic diagnostics (geneticsandsociety.org). They warn that such results “only assess risk” and shouldn’t be over-interpreted. Unfortunately, marketing messages can drown out these nuances. Some companies use flashy visuals and promises like “find out if you’re genetically prone to weight gain or success in athletics” – ventures beyond health into dubious trait prediction. There’s also the issue that many consumers of DTC tests are of diverse ancestries that the scores weren’t calibrated for, leading to less accurate results (a fact usually buried in fine print).

In short, the consumer-facing side of PRS is a Wild West right now: exciting, but full of snake oil potential. Regulators like the FDA have begun to scrutinize these, but the global nature of the DTC market makes enforcement hard. For responsible parties, the focus is on education – ensuring that if people get these scores, they understand the limitations, the probabilistic nature, and the continued importance of non-genetic factors. As one commentary put it, “polygenic scores provide probabilities, not prophecies” – a distinction that laypeople can easily miss amid hype.

Psychosocial and Clinical Workflow Challenges

Even if we solve technical issues and prove utility, implementing PRS in healthcare on a broad scale will face practical hurdles. Genetic literacy among physicians is variable – many have not been trained in interpreting polygenic risk. Already, reports from companies like Myriad (for RiskScore) need to be accompanied by education for providers so they don’t mis-advise patients. There’s also time constraints: discussing a PRS result properly might take a chunk of an appointment, something many busy clinics can’t spare, especially if the significance is marginal. Healthcare systems will need to decide who delivers and manages this information – primary care docs? Genetic counselors?

AI-driven reports directly to patients with chatbots? None of those is a perfect solution, and scaling genetic counseling to millions of people is impractical given current workforce. So there’s a delivery challenge: how to integrate this data in a seamless, non-disruptive way. Additionally, one must consider the psychological impact on patients. Some might experience anxiety or distress from learning they have elevated genetic risk.

Others might misremember or miscommunicate it (“I have the cancer gene” someone might erroneously say after finding out a high PRS, confusing it with a deterministic mutation). Ensuring proper understanding is an ongoing service requirement, not a one-off. All this introduces costs – both monetary and in terms of clinician workload. And speaking of costs: while genotyping is cheap (under $50 now for a genome-wide array) and often one test can compute scores for many diseases, there’s the question of who pays. Insurance might not cover a PRS test if it’s deemed not standard-of-care yet.

Some forward-looking healthcare systems (like in UK’s NHS pilot projects) are absorbing that cost in research contexts (genomicseducation.hee.nhs.uk), but broader routine use will need cost-benefit justifications. If an insurer doesn’t see evidence that PRS improves outcomes, they won’t pay for it. In fact, some insurers have explicitly labelled polygenic tests as “experimental/investigational” and not reimbursable so far. The reimbursement landscape is evolving: one can imagine in a few years certain PRS (e.g. breast cancer) might be bundled into covered services if guidelines endorse it. But until then, it could remain an out-of-pocket luxury or curiosity for many.

Noise, Updates, and Reproducibility

Another pragmatic limitation is that polygenic scores are a moving target. As new GWAS are published, the “best” PRS for a given disease might change (new variants added, better weights computed, etc.). This is good scientifically (improving accuracy), but it means a PRS you got last year might be “Version 1.0” and now there’s a Version 2.0 that reclassified you. If a patient has a score and then later it’s updated, how do we handle that? It’s not like a blood test which is measured anew – PRS is derived from static DNA, but the interpretation can change as knowledge evolves. There may need to be mechanisms for updating genetic risk reports, which is not something our healthcare IT systems are currently built for.

It’s akin to software updates for your genome risk profile. Furthermore, there are different methods and companies producing scores – and they don’t always agree. One might tell you you’re top 10%, another says top 20%, depending on training data and algorithms. This lack of standardization can cause confusion. The field is moving toward more open sharing (like the PGS Catalog) to reduce discrepancies, but for now, reproducibility can be an issue when scores are derived in silos.

Despite these caveats, none of the limitations are show-stoppers in the long run – they are challenges to be managed. Genetic diversity gaps can be filled with more data and better methods (though it’s a slow process). Clinical utility can be demonstrated through rigorous studies and pilot implementations leading to refined guidelines (genomicseducation.hee.nhs.uk). Ethical misuse can be curtailed with regulation and professional standards (as we see ACMG doing for embryo selection).

Education can improve for both doctors and patients. The key is ensuring we don’t rush headlong into using PRS in high-stakes settings without laying the groundwork. A sober approach acknowledges these limitations and works on them, rather than being blinded by the hype.

To give a concrete illustration of the current state, a recent expert consensus in the UK for breast cancer said: yes, we can calculate PRS and it stratifies risk, but we recommend using it only in very specific scenarios for now (like refining risk for women on the cusp of clinical decision thresholds), and only with proper counselingmdpi.com. That encapsulates the prudent stance: promising tool, handle with care.

Policy, Reimbursement, and the Business of PRS

Where science meets society, polygenic scores have stirred discussions in policy circles and attracted significant investment in the private sector. It’s worth examining how regulators, healthcare systems, and businesses are approaching PRS in 2025, as this context will shape how quickly (or slowly) these scores transition from the lab to everyday life.

On the policy and regulatory front, as mentioned, professional bodies like ACMG (American College of Medical Genetics) have released statements to guide the clinical application of PRSgeneticsandsociety.org. The tone of such statements is cautious. They emphasize points we’ve discussed – that PRS are not diagnostic, that a “low risk” result doesn’t guarantee safety, that evidence for using PRS to guide care is limited so far (geneticsandsociety.orggeneticsandsociety.org).

By doing so, they set a tempo for clinical uptake: essentially urging that PRS should currently be used in research or in a managed context, rather than indiscriminately.

Another example: the American Heart Association’s scientific statement (2022) did a comprehensive review of PRS for heart disease and concluded that while PRS can modestly improve risk prediction, more studies are needed to see if acting on PRS actually improves patient outcomes (ahajournals.orgacademic.oup.com).

They pointed out issues of ancestry and urged including PRS in research but stopped short of recommending it for widespread clinical use. Similarly, in Europe, agencies like the ESHG (European Society of Human Genetics) and national genomics initiatives are analyzing how PRS might fit into public health. The UK’s National Health Service has run pilot programs (like the heart disease PRS pilot) but these come with academic oversight and evaluation components to decide on scaling (genomicseducation.hee.nhs.uk).

One policy area to highlight is screening guidelines. For example, an international consortium is working on personalized breast cancer screening guidelines that incorporate PRS, breast density, and other factorsecancer.org. The idea is that in a few years, countries might officially endorse risk-based screening intervals for breast cancer (so high-risk get screened earlier/more often, low-risk later/less). If that happens, it would be one of the first formal incorporations of PRS into standard preventive care. It’s a tricky policy balance: they have to be convinced that tailoring by PRS does more good than harm and is cost-effective.

The preliminary data from trials like WISDOM (in the US) and MyPeBS (in Europe) will inform that. In other disease areas, such as colorectal cancer, policy folks are keen but waiting on more evidence before saying “yes, do colonoscopies based on PRS-stratified age schedules.” So, policy is cautiously supportive but evidence-bound – a healthy stance.

In terms of reimbursement, currently most insurers do not pay for polygenic score tests as a standalone. If you go through a doctor to get one, it might be billed as a panel or part of a broader genetic test, and likely denied coverage unless under specific conditions (e.g. some clinics might offer it as part of research or a package). The exceptions are scenarios like Myriad’s RiskScore, which is bundled with their myRisk hereditary cancer panel – when they bill insurance for the panel, RiskScore comes “free” (Myriad absorbs the cost).

Essentially they found a business model where the PRS ups the value of their existing test without needing separate reimbursement. But for, say, a primary care physician wanting to order a polygenic risk test for heart disease from a lab, insurance would almost certainly label it experimental.

We’ve seen early attempts by companies to get CPT codes or reimbursement for PRS-based tests (like one company might try to create a “comprehensive risk assessment” product), but it’s an uphill battle until guidelines back it. Medicare and big payers tend to follow guidelines, so if USPSTF or professional societies don’t recommend something, they won’t pay. However, if and when guidelines do incorporate PRS (say in breast cancer), you can expect the payers to come around, grudgingly perhaps, when they see it could prevent later expensive cancers by catching them early.

On the venture and business side, PRS has been hot. Investors love the buzzwords of “AI-driven genomics” and “preventive personalized medicine,” and polygenic scores sit right at that intersection. In the past 5–10 years, multiple startups have popped up focusing on polygenic risk analysis and its applications. For example: Genomics plc (UK-based, co-founded by leading geneticists) has raised tens of millions of dollars to develop integrated risk scores for common diseases and work with healthcare systems to implement them.

They ran the pilot with NHS we discussed, and recently reported outcomes where adding their genomic risk tool could identify something like 1 in 100 people who’d benefit from statins that current practice would miss (biocentury.com). They position themselves as a kind of analytics company helping health systems triage risk.

Another category is companies like Allelica and Polygenic Health which offer PRS calculation platforms for clinical labs or hospitals – basically enabling others to compute scores from genotype data. They have raised funds on the premise that soon many will need PRS computation capacity. Then there are consumer-facing startups: We mentioned Orchid Health (which initially offered couples a report on their future kids’ risk and now also embryo testing) – they got media attention and VC backing with the bold promise of “empowering parents to reduce their child’s disease risk”freethink.com.

Similarly, GenePlaza, Impute.me, and other online platforms let people upload data and get myriad polygenic scores (outside FDA purview if they label them as informational). 23andMe itself went public in 2021, riding general enthusiasm for consumer genetics, and part of its growth story involves expanding the health reports (including more polygenic ones presumably) for its ~12 million customers.

Large diagnostics companies have also jumped in: Myriad, as we saw, integrated PRS into its product line to differentiate itself (cancernetwork.com). Laboratory Corporation of America (LabCorp) acquired a small company that was developing PRS for prostate cancer screening. And major genome-sequencing firms talk about adding PRS analysis as an extra for their whole-genome sequencing reports.

What’s the business model? For now, mostly either DTC sales (consumer pays out of pocket for a test/report) or B2B integration (selling the service to clinics or labs who bundle it). Some companies bet that once PRS becomes mainstream, insurance will cover it like any other lab test, and they want to be the established player by then. Indeed, one can envision a future where at a certain age (say 18 or at birth), everyone gets genotyped and multiple PRSs are part of their health record.

That’s a moonshot concept – it would require significant evidence and policy change – but if it happened, it’s a multi-billion dollar market in genotyping and data analytics (which is why companies are vying for position now). In countries with single-payer systems, it could be rolled out centrally (the UK’s NHS for instance is currently collecting half a million whole genomes through the Our Future Health program, explicitly planning to use risk scores in the feedback to participantsourfuturehealth.org.uk). In the US, it’d likely be more piecemeal via commercial offerings unless a consortium or large insurer decides to push it.

Another area of venture interest is pharma: drug developers see polygenic scores as a way to define patient subgroups. For example, a company with a preventive Alzheimer’s drug might collaborate with a genomics company to identify high-risk candidates for a trial, as recruiting those with high PRS can make a trial more likely to succeed (enriching for those who will progress). If that drug works, they might also market it as “particularly indicated for high-genetic-risk individuals” – opening a niche to prescribe a drug based on genetic risk (similar to how statins are particularly justified in familial hypercholesterolemia, a monogenic condition, one could imagine a moderate risk drug being justified in those with polygenic hyper-risk).

We’re not quite there yet in practice, but pharma is absolutely paying attention to PRS. You can find papers and conference talks from industry scientists about how PRS can help “identify unmet need populations” (read: people who are currently not labeled high-risk by guidelines but genetically are at high risk – which could be a market for prophylactic therapies).

Finally, we must mention data privacy and ownership. Genomic data is valuable, and when you get into multi-disease risk profiling, it raises privacy questions. If an employer somehow accessed an employee’s polygenic risk for e.g. mental illness, could they subtly discriminate? Or if an insurance company got a hold of a risk profile, could they hike premiums or deny coverage? (Laws like GINA were meant to prevent this in health insurance, but life insurance is another matter.) As more companies build databases of genomes linked to PRS and health records, ensuring that data is not misused or breached is a policy concern.

In the wake of these issues, some have even proposed that polygenic scores be considered sensitive personal data warranting special protection, akin to one’s entire genome. There’s also a flipside: if PRS becomes medical information, people might have a right to know it (one could argue a physician has a duty to inform a patient of significant risk factors). We’re in early days of legal frameworks around this.

In summary, the investment and strategic interest in polygenic risk is high, but it’s tempered by the realization that clinical adoption will be incremental. The path to monetization runs through proving clinical value. So businesses are often pairing with healthcare providers or doing pilot projects to gather real-world utility data. Those who can show that using PRS saves money (e.g. by preventing expensive diseases) or improves outcomes will have the best case to make to payers and policy makers. Meanwhile, regulators are watching closely to avoid hype from overtaking evidence.

One can detect a cautious optimism: nobody wants to miss out on a genuinely useful medical innovation, but neither do they want a redux of the “genomic hype” of early 2000s where promises fell flat. The next few years, as results from ongoing trials and pilots come in, will likely determine how quickly PRS moves from niche to normal in healthcare.

Conclusion: Polygenic Score or Overly Quantified Proxy?

After surveying the landscape – the science, the hopes, the doubts – we circle back to the contrarian question: Is the polygenic risk score just another overly quantified proxy, more sizzle than steak? In other words, are we attaching perhaps undue significance to a fancy number that ultimately serves as a stand-in for things we largely already knew?

What a PRS tells us could often be approximated by a good family history and awareness of known risk factors. Indeed, a high PRS for, say, type 2 diabetes will very often correlate with having multiple family members with diabetes (which any doctor would already act on) and with being of an ethnicity that has higher prevalence.

A PRS for heart disease captures aspects of family history, cholesterol metabolism, blood pressure regulation, etc., much of which can be partly observed through phenotype. So one might cynically say: PRS is a high-tech way of quantifying familial and inherited risk that we always suspected was there. It gives a number to what was qualitatively known.

But does that fundamentally change management or outcomes? In some cases, maybe not much. It’s like having a more precise thermometer for risk – useful, yes, but if you already knew a patient was high risk due to strong family history and suboptimal lifestyle, the genetic score might only confirm the obvious.

On the other hand, there are scenarios where PRS genuinely adds new information – identifying someone with hidden risk despite unremarkable family history (perhaps due to smaller contributions from many distant relatives, etc.), or revealing someone not at risk despite a concerning family pedigree (maybe the family’s cases were due to bad luck/environment and the person didn’t inherit the risk variants). Time will tell how often these scenarios occur and matter.

A contrarian might also highlight how the PRS boom echoes previous “revolutionary” predictive metrics that ended up with limited utility. We’ve seen waves of enthusiasm in medicine for various biomarkers and algorithms. Remember when CRP (C-reactive protein) was touted as the key to heart disease prediction? Or when gene expression tests for disease risk were in vogue? They each found a niche, but largely didn’t upend practice as expected. There’s a possibility that PRS, after the hype settles, will be one tool among many – helpful in specific cases, but not a paradigm shift across the board.

Critically, outcomes are what matter. If widespread PRS use doesn’t demonstrably improve health outcomes (lives saved, diseases prevented, cost reduced), it will remain more of an academic plaything or a DTC curiosity. The burden of proof is on the genomics community to show that this deluge of data can be translated into healthier patients and more efficient healthcare. Otherwise, PRS could indeed go down as an “overly quantified proxy” – a sophisticated measure that, in practice, didn’t change the game significantly.

It’s worth noting that even the strongest advocates temper their claims now. They no longer say “this will predict disease,” but rather “stratify risk” and “enable targeted prevention.” That is a realistic framing. It acknowledges probabilistic nature and the need for complementary factors. Polygenic scores are a proxy for genetic predisposition, and genetic predisposition is itself only one proxy for overall disease risk (environment and behavior being huge parts of the picture). So we’re dealing with proxies of proxies. If one quantifies a proxy to more decimal places, it doesn’t necessarily yield commensurate real-world improvement – sometimes yes, sometimes no. It’s a classic diminishing returns problem in predictive analytics.

In a reflective vein, one might consider the philosophical aspect: Our modern medical system loves numbers. We measure cholesterol to one decimal, blood pressure to the exact mmHg, risk scores to percentages. PRS slots perfectly into this quantification mindset – it gives an illusion of precision (“your risk is in the 87th percentile!”). But a dry observer could question whether this fosters a false sense of certainty in what are, end of the day, messy human outcomes. The danger of over-quantification is that it can distract from qualitative factors that are hard to measure but very important (like social determinants of health, or plain old common sense lifestyle advice). We must ensure PRS augments, not eclipses, holistic patient care.

So, is PRS just an overly quantified proxy? The answer likely lies in the middle. It is a proxy – a fancy summary of genetic variants, many of which we don’t fully understand mechanistically. By itself, it’s not the disease or the destiny. It’s an index. But it’s a proxy that in some contexts carries useful signal that wasn’t readily available before. If used wisely, it can refine our approximations of risk. If used poorly or over-sold, it can mislead and overcomplicate.

The polygenic risk score, in trying to foretell our medical future, may chiefly be telling us something about ourselves in the present – our endless fascination with prediction and our hope that more data can reduce life’s uncertainties. It’s a worthy endeavor, to be sure. Yet one should prefer not to mistake the map for the territory.

A PRS is a map of genetic propensity; life’s terrain includes environment, chance, and choice which no score can fully capture. In the end, the utility of polygenic risk scores will be proven not in how well they rank risk on paper, but in whether acting on them leads to better outcomes in the real world. Until that evidence is incontrovertible, a measure of skepticism remains the prudent companion to any excitement.

Congress famously once asked for a “weather report” on the genome; the PRS is perhaps a localized forecast – occasionally helpful, occasionally a false alarm, and always subject to change. It’s an impressive quantitative feat born of big data and statistical wizardry. It’s not magic. In the final analysis, if PRS lives up to even part of its promise, it will become a valued piece of the medical puzzle. If not, it will join the graveyard of buzzwords past. For now, consider it a tool under construction – potentially powerful, undeniably interesting, but not a crystal ball. In other words, a valuable proxy when kept in perspective, and an overly quantified one when not.

Whether PRS is revolutionary or just revelatory (of what we implicitly knew) will be decided in the coming years. But one thing is certain: it has reinvigorated the conversation about personalized medicine, forcing us to grapple with how much we want to know about our genetic fortunes and what we’ll do about it. And in that sense, regardless of outcome, polygenic risk scores have already made an indelible mark on the discourse – a dry irony, perhaps, that a simple weighted sum of SNPs could generate so much passionate complexity.