'Some of them have accuracy that's close to zero': Experts unpack the promise and pitfalls of genetic tests aimed at consumers

The past decade ushered in a surge of discovery in the field of human genetics — and simultaneously, more genetic technologies made their way out of the lab and into the consumer marketplace.

This tech includes at-home genetic tests for learning about health risks and ancestry, as well as polygenic embryo selection, which enables prospective parents undergoing in vitro fertilization (IVF) to predict the future traits of the resulting embryos. If these products work as advertised, they could improve health outcomes; but are they really as powerful as their marketing claims? And what impact could these technologies have on society if they're used irresponsibly?

In a new book, "What We Inherit: How New Technologies and Old Myths Are Shaping Our Genomic Future" (Princeton University Press, 2026), bioethicist Daphne Martschenko and sociologist Sam Trejo unpack persistent myths about genes that shape scientists' and the public's views of these technologies. They weigh up the possible pros and cons of using such tools, ultimately concluding that the field is in desperate need of regulation. They argue that allowing it to advance without guardrails could deepen existing social inequalities.

Live Science spoke with Martschenko and Trejo about their book and perspectives on genomic technologies, especially those aimed at consumers.

Nicoletta Lanese: What prompted you to write this book now?

Daphne Martschenko: We realized we were both similarly disillusioned by the bitter academic debate that was going on over whether and how to do genomic research on behaviors and social outcomes — this field of "social genomics" that we focus on in the book.

We also had a goal of wanting the public to understand these consumer products that are coming out — direct-to-consumer genetic testing, polygenic embryo selection. [We wanted people to] understand some of the science behind them and the limitations behind them so that they can make informed decisions when thinking about whether or not to access those technologies.

Sam Trejo: A lot of it has to do with [the fact that] we're pretty early on in this "post-genomic" era. We had the first genome sequence about 25 years ago, but it's only really been in the last 10, maybe 15, years that we've started to get large enough genomic databases to make rigorous discoveries — because the genome is so big, and each particular region of it actually has, it turns out, a very small contribution to most traits.

Now that the science is improving, our ability to take a person's genome — have them spit in a cup, process that information, and make predictions about a wide range of traits — is growing. It could be from how tall someone is going to be, how far they're going to go in school, their likelihood of developing Crohn's disease or schizophrenia. Over time, our ability to summarize a person's genetic predisposition has gotten better, and it's increasingly used in scientific research in a wide variety of ways.

But there are these lingering questions about, to what extent should we take these technologies that were designed for research and actually use them in the wider world?

NL: In the book, you address myths about genes and misconceptions around the influence of "nature versus nurture." Why was that important?

ST: In the book, we call this [the] "destiny myth," which summarizes the idea that a person's DNA affects their traits, their diseases, their life outcomes in simple, straightforward, immutable ways that are biological and distinct from the social and cultural aspects of their life — the idea that DNA is destiny, that if you have a genetic predisposition for something, then there's little that we can do about it.

What we try to do in that chapter is really drill down to, where are the origins of these misconceptions that we have about DNA? And these recent genomic discoveries that say, "Oh, we've identified the regions of the genome that are relevant for educational attainment or that are relevant for depression" — what do those actually mean?

The big story is, even though we're starting to identify many regions of the genome that correlate with, or even causally affect, a wide range of medical traits, social traits — even though that's true, we still don't really know why. We don't know the mechanisms that connect DNA differences to differences in people's life outcomes.

DM: When we talk about some of these downstream applications, like the use of polygenic scores in embryo selection, we discussed how the polygenic scores are a "black box predictor." [Editor's note: Polygenic scores predict the likelihood that a given trait or disease will emerge based on an individual's genome. Some companies offer polygenic scores for IVF embryos, enabling parents to select embryos with higher scores for desired traits.]

When Sam was saying that there's poor understanding of what causal mechanisms are, we don't know why a genome-wide association study flags variants that are associated with whatever the trait is that the researchers are looking at.

NL: Where might it be appropriate to use polygenic scores for health applications?

DM: I think there's more of an appetite for using polygenic scores for medical conditions like heart disease or type 2 diabetes. That's less controversial than, for example, using polygenic scores in something like embryo selection for intelligence, or offering direct-to-consumer genetic testing for a trait like intelligence.

Part of that is because of the long-standing history in which claims regarding genetic differences in socially valued traits, like intelligence, have been used for social harm. That's something that we talked through in the book: how the destiny myth and the "race myth" have been used to justify laws outlawing interracial marriage or legalize involuntary sterilization. [Editor's note: The authors define the "race myth" as the false belief that DNA differences divide humans into discrete and biologically distinct racial groups.]

ST: In the book, we talk about "application genetic screening," which would be used to stratify access to certain medical interventions or treatments based on polygenic scores. So, if you go to the cardiologist, they're going to assess your risk of having a heart attack in the next few years. Having a high genetic predisposition could be something that gets indexed by a polygenic score. That could be something that people are comfortable with doctors using to decide "What level of statins should I prescribe, or what other interventions are appropriate here?"

There's this idea that it's [polygenic scoring is] going to allow us to better target our resources to the people who need them the most. We're identifying this hidden risk for heart disease or these other negative traits that we can then help ameliorate. On balance, the use of polygenic scores in the clinic in this way could reduce the differences in outcomes between somebody who has a high risk for heart disease and a low risk for heart disease.

The flip side would be something like using polygenic scores for private school admissions. As far as Daphne and I can tell — or the experts that we've talked to — there's no kind of legislation that would prevent a private school from considering polygenic scores in addition to, you know, personal essays, past academic performance, in determining how to admit students. This isn't something that we're actually seeing in the world, but as a hypothetical application, I think this is something that would make many people uncomfortable.

NL: What would you flag as limitations of direct-to-consumer genetic tests?

DM: There are a number of companies that offer genetic tests for a wide range of traits. If you can imagine a trait, there's probably a company somewhere that is offering a genetic test for that. Things like facial attractiveness, athletic ability, political views, intelligence, heart health, brain health — you name it, there's a company that's trying to sell it to consumers.

One of the things that we point out is how companies offering these kinds of tests sometimes use the destiny myth to market the product they're selling. They overstate the role and relevance of DNA to make the consumers feel that this is really important information for them to have. We, in the book, debunk the destiny myth.

ST: There's also not a lot of transparency from the companies in terms of what datasets they are drawing upon, [or] how they analyze the samples that people are sending in, in order to spit out the genetic report that the consumer has purchased.

For most of the traits that people are interested in, aside from very specific diseases and disorders — Huntington's disease, cystic fibrosis, Tay-Sachs [disease], sickle cell [disease] — most human traits are polygenic, which is to say that many, many, many regions of the genome matter for shaping that trait. Sometimes, these direct-to-consumer tests will tell somebody that they have a very high genetic risk for some negative outcome, but they're only looking at three variants, when actually 10,000 variants matter.

So there's not clear information on these direct-to-consumer sites about how predictive or how accurate these tests are. And certainly some of them have accuracy that's close to zero [for specific traits] but are provided nonetheless.

NL: And what are the limitations of polygenic embryo selection?

ST: What polygenic embryo selection does is use these scores to try and change the expected characteristics of the child that the set of prospective parents will have. Before we decide which of these [IVF] embryos is actually going to be implanted and become a fully realized human child, we're going to genotype all of them. We're going to see which DNA they happen to inherit from each parent, and we're going to pick the embryo that we think has the "best" or "healthiest" genetic characteristics.

But importantly, how effective this technology is at present is limited by a lot of things. … The truth is, for most traits right now, its accuracy is quite limited.

Height [as a selectable trait] has two things going for it: One, we have very large sample sizes to train our polygenic scores on, partly because height is something that everybody has and it's easy to measure. And height is very heritable; it's very genetically influenced. In the U.S., about 80% of the variation in height is due to people's DNA, so it's kind of the best-case scenario. But some traits are much less genetically influenced.

Another important piece of the puzzle is the number of embryos that you're able to select from. If you're only selecting from a few embryos, then even if you pick the one that has the highest polygenic score for a particular trait, you're actually not going to get that big of a change, on average.

The technology relies on our ability to identify the regions of the genome that matter for a given trait … and polygenetic scores are trained on sample sizes of, ideally, millions of people. Typically, though, those people are from one particular region of the [human] family tree: the European ancestry region.

There's very limited portability of these products to other ancestries. The accuracy declines for Hispanic Americans, Asian Americans, Black Americans, who tend to have ancestry from different regions of this big family tree of humanity.

DM: I want to make clear that the "big family tree of humanity," and the genetic ancestry that researchers crudely draw over that family tree, is not synonymous with race. … The social process of race is where we look at physical, outward appearances of folks and make decisions about how we treat them and understand them. Race is a social process that's about power, and it's not the same thing as the great family tree of humanity.

NL: What are the big takeaways from this book?

DM: I would say, to the academic researchers who are enthusiastic about polygenic scores and how they might be deployed in the world, and for those who are more cautious and apprehensive, our message is that if we want to ensure that these technologies are used in ways that maximize good and minimize harms, it's very important that we take the time to really listen to each other and understand why we're disagreeing.

Something that Sam and I learned is that we didn't have to agree on everything in order to agree about the need for regulation of these technologies and to develop a preliminary framework for thinking about how we might go about that regulation.

In the last part of the book, we think about regulating the use of polygenic scores in not just embryo selection but also direct-to-consumer genetic testing and screening in settings like schools or financial lending. [On that front] we also have a message for policymakers, really calling for the need for greater regulation of this technology and offering a potential path forward for at least getting the conversation started.

For members of the public, a key goal is to help folks understand, when they go to a company like 23andMe or Ancestry and get their ancestry results, what is the information that they're receiving? How are these companies generating these tests, and what do I need to know so that I understand what the limitations of them are?

When it comes to the consumer products that are related to the social behavioral traits or to embryo selection, it's also about helping people understand, what is the science or lack of science behind some of these products — so again, consumers can make informed decisions about whether they want to spend their money to buy a genetic test for something like facial attractiveness or math ability, understanding the dearth of scientific evidence to support those kinds of consumer tests.

Editor's note: This interview has been lightly edited for length and clarity.