Interview with 23andMe: How surveys help revolutionize genetic research

What if the right survey question could help you find a cure?

Hyperemesis gravidarum (HG) or extreme morning sickness, was described by one patient as, “nausea so intense and all-consuming you feel like you’ve been poisoned.” Duchess Kate Middleton was hospitalized with it in 2018, and Jane Eyre author Charlotte Bronte is believed to have died from it. And yet until recently many people—including some doctors—wouldn’t have considered it a serious condition.

Why? Because we don’t know much about the cause. But leading consumer genetics and research company 23andMe was able to use survey research, plus their database of genetic data, to uncover an important new clue: genetic associations between two genes—GDF15 and IGFBP7—and people who reported morning sickness.

The new study might change the way researchers look for new treatments. Per 23andMe, “Up until now much of the focus of research around HG has been on hormonal fluctuations in the body. This study found no evidence for that link, and instead suggests that the expression of the gene involved in placenta formation may play more of a role.”

This breakthrough,and many of the 139 other pieces of scientific research that 23andMe has published since 2010, was largely a result of the company’s innovative way of marrying survey data from its consented research participants with its pre-existing genetic database.

To understand more about how 23andMe is using surveys to innovate in the field of medical research, we talked to one of their data scientists, Olga Sazonova, Ph.D.

We asked Sazonova to provide a bit of background on their research program.

Here’s how it works. You may know that 23andMe customers send in a DNA sample (saliva) for analysis. Using that sample, 23andMe is able to study that person’s genome and provide reports about their genetic health risks, traits, wellness, and more, in addition to the customer’s ethnic background.

Next, they send surveys to customers who have opted into participating in research. Out of the 23andMe customer database, more than 80 percent of customers do consent to participating in research. 23andMe’s surveys are shared through their customer dashboard and promoted via email. Customers can choose to answer as many or as few as they like in order to contribute to the research program, if they’ve opted in.

(Note that 23andMe onlyuses data from customers who explicitly opt in. And even if you do opt in, your data is still de-identified and if shared with an academic or industry collaborator -- is only shared in aggregate. They’ve also worked with an Institutional Review Board (IRB) —an independent committee that evaluates protocols and ensures researchers are following ethical research practices—since 2010.)* This IRB is an independent administrative body composed of scientists, non-scientists and community members.

When the survey responses are in, 23andMe combines the survey data with the genetic data to uncover new insights. Do people who report to having a cancer diagnosis have certain genetic variants  in common? Are people with a certain version of a gene more likely to struggle with insomnia? On average, respondents answer 250-300 questions, which gives 23andMe researchers enough context to uncover powerful new insights.

23andMe collaborates with both academic and medical institutions to plan studies and publish new research.

“Survey data and genetic data are the two pillars on which we base all of our studies. It’s because of these two unique data sets that we’ve been able to pioneer a new way of doing research.”

23andMe's novel approach of combining survey and genetic data could create a new paradigm for how research is done, according to NBC’s profile of the company: “To complete traditional medical studies, researchers must first assemble a group of people with a particular disease or trait, and then search for genetic clues. But 23andMe can take the reverse approach—selecting the genes to study based on the reported health conditions among its existing data pool.”

Sazonova says 23andMe has proven this approach can work. They've found real associations between the self-reported information from their surveys and the genetic traits present in the groups they're studying.

And as it turns out, 23andMe’s data—often based only on survey responses and genetic information—is just as trustworthy as the findings from clinical research. Here’s how Sazonova puts it:

“Our findings are just as good, and in many cases, they’re better. When I say our research is ‘just as good,’ I mean it replicates findings from clinical studies. And when I say ‘better’, I mean we then find new associations, that others haven’t been able to find.”

Why does 23andMe have such confidence in its results, especially without medical professionals to formally examine and diagnose the people in its studies?

What it boils down to is scale. Whereas traditional clinical studiescan be in-depth but heavily limited in sample, 23andMe’s trove of genetic and survey data can offer much higher numbers. 23andMe’s massive community of research participants are all potential candidates for new research, and all they have to do to participate is opt in and answer a few questions.

According to Sazonova, 23andMe studies regularly have 5 or even 10 times more participants than you’d find in traditional genetic studies—one of the advantages of using online surveys instead of requiring in-person contact.

Combining survey data and genetic data isn’t only important for the advancement of medical and scientific research. It also has a real impact on 23andMe’s product—the online dashboard that enables 23andMe customers to learn about their genetics.

23andMe generates dozens of reports that continue to give customers more insights about their DNA long after their initial purchase—and each report is a new source of value. Here are a few examples of 23andMe reports that were developed using survey data from their consented research participants:

As 23andMe’s research team uncovers new genetic associations, they’re able to provide more of those reports—which in turn make for happier customers.

Sazonova says,”When 23andMe first started, our ability to deliver reports for our customers was restricted to only research that had already been done by others. Once we built up our research program, we had a new mechanism with which we could engage with our customers, give them this opportunity to actively participate in new breakthroughs, and generate discoveries ourselves.”For example, 23andMe released a Type 2 Diabetes Health Predisposition report this year, based on their survey data and genetic research. For this report, the 23andMe team created a statistical model using real-world data from more than 2.5 million 23andMe customers who opted into research. Some of those customers have type 2 diabetes, and some do not, but together the data has allowed 23andMe to create what’s called a “Polygenic Risk Score”.

Polygenic Risk Score means that 23andMe makes a determination about your risk based on observing how a number of different genetic variants behave together, rather than just one or a couple. By holistically assessing patterns of genetic variation that are typical for people with Type 2 Diabetes, 23andMe is now able to tell you whether you might be at risk.

Survey data enabled the researchers to arrive at findings which then became a powerful part of the company’s offering for its consumers. Sazonova added that, “We’re hoping to continue to use these kinds of polygenic modeled reports to inform our customers of their health predispositions. The data our customers contribute through our surveys allow us to better understand the genetic underpinnings of various diseases, some of which affect millions of people’s lives.”

23andMe is also now the first and only direct-to-consumer company with FDA authorization to tell people about diseases that they’re at risk for. These reports are based on genetic information. But people who participate in the survey studies can see how their results compare to others who have taken the surveys—which might also illuminate something about their own condition.

The research team at 23andMe does both confirmatory (checking a hypothesis) and exploratory (open-ended) research.

Sazonova explained, “We don’t want to limit ourselves to areas where we already have a good reason to believe that genetics matter. We want to open ourselves up to all sorts of unexpected discoveries about everything—and that list is probably beyond what we currently have imagined”

In 2012, 23andMe discovered the genetic variant behind cilantro aversion (which is when certain people eat cilantro and think it tastes like soap). Cilantro aversion might seem like more of a question of preference than of genetics, but it appears to be coded into our DNA.

But some of 23andMe’s research might not even involve genetics at all. In some cases, they might be able to rely on only survey data to arrive at new findings.

23andMe researchers can also examine how different phenotypes (anything about a person that isn’t strictly genetic) affect one another. This type of research looks at how different types of people—based on their survey responses—might act in different ways. Are people who like hip hop more likely to travel outside the U.S.? Are people with red hair more likely to enjoy dancing?

Sazonova offered one example using a computer-based  cognitive game used to assess people’s risk aversion. It involves virtually inflating balloons until they pop - players keep points for every inflated and unpopped balloon but it’s impossible to know which balloon will pop and when.. The game closely mimics  a well-established tool in psychological research called the Balloon Analogue Risk Test (BART).

23andMe compared participants’ results with a wide range of survey responses and looked for factors that correlate with risk taking or risk aversion.  The researchers found that risk-prone people are less likely to describe their desk as “neat and orderly”—and (perhaps unsurprisingly) are more likely to report being in car accidents.

The takeaway? Even if you don’t happen to have access to a massive genetic database, , you can still find compelling insights from your survey research.

Of course, not every company has access to a database of genetic information from millions of enthusiastic research volunteers, but most have a dataset of some kind which—when enriched with survey data—can reveal a fascinating and totally unique set of insights.

Each survey question that 23andMe research participants answer is a new layer of context for understanding how people work at a genetic level and beyond. 23andMe can slice and dice their data in innumerable different ways to discover new findings.

And they’re making the most of it. The company has contributed to bodies of research on medical conditions from Parkinson’s disease to insomnia to cancer. In the Information Age, companies that think creatively about how to combine data in different ways are the ones that innovate best, grow fastest, and ultimately change the world.

*People who choose to participate in the research program can also see exactly which studies their data helped support through their dashboard—23andMe explains more here.