Medical Examiner

What’s in Your Genes?

Some companies analyzing your DNA rely on junk science.

DNA strands.
At 3 billion bases in length, the genome is massive, so user interpretation can easily go awry. 

Image by Dynamic Graphics/Thinkstock

Each person is born with an assortment of genes that raise or lower his or her risk of certain diseases. Some genetic variants help you live longer; others make you vulnerable to Alzheimer’s. Genetic testing services such as 23andMe launched with the promise of informing consumers about the risk factors latent in their genes. But these services failed to get approval as medical diagnostic tests from the Food and Drug Administration, and in 2011 the FDA prevented 23andMe (and similar services) from providing information about the health implications associated with most genetic variants.

Despite that roadblock, I set out to get my genome tested in order to learn more about my own health. I have a Ph.D. in genetics, and I reasoned that I could draw on my knowledge of and access to the scientific literature to diagnose my own genetic risks. I quickly discovered that the FDA hasn’t actually prevented consumers from getting access to the health-related information within their genomes. Instead, a spectrum of bewildering and even dishonest sources has sprung up, some of which rely on junk science.

The first website I ran across in my hunt to understand my own genome is called SNPedia. As its name suggests, SNPedia is a Wikipedia-style repository of knowledge but specializing in the health information associated with the sort of genetic variants 23andMe tests for. (The technical name for this kind of variant is “single nucleotide polymorphism,” which provides the site its prefix.) SNPedia turned out to be the database for a product called Promethease, which offered to connect my genetic variants to each record in SNPedia’s voluminous database for the price of only $5.

I jumped at the chance. My first foray into the database led me to genetic variants associated with long life, and one of the first I came across was this one. As I followed the link to the associated paper, I realized that the article claiming it increased lifespan had been retracted, or formally declared incorrect, in July of 2011.

Promethease claims that it generates a staggering 100 to 500 health reports a day. For each of these, a consumer uploads his or her genome as sequenced by 23andMe or another provider, and Promethease spits out a catalog of links to genetic risk factors. That number is on par with the average daily number of genetic test kits 23andMe has processed over its history. A further 2,000 to 3,000 people visit SNPedia per day, and many more draw on the Promethease database through other companies that have licensed it and offer health interpretation services. Together, Promethease and SNPedia appear to be providing health information to a substantial fraction of the consumers getting their genomes tested.

In theory, SNPedia is a democratization of science: Like other Wiki-style databases, anyone can contribute to any article, regardless of his or her qualifications. As co-founder Mike Cariaso wrote, “You are judged by the quality of your work, not the degrees you hold.” In practice, Cariaso, co-founder Greg Lennon, and an army of Internet-crawling bots create and edit most of the articles, harvesting variants from the latest issues of scientific journals. Of the past 1,000 edits, only about 60 came from independent contributors’ accounts. Cariaso controls the quality of SNPedia articles rather strictly, and he was able to point me to several cases in which he modified or deleted pages when scientific articles were criticized or withdrawn.

Despite Cariaso’s careful attempts at quality control, a subreddit associated with SNPedia is filled with tales of confusion and panic. These stories may not be representative of SNPedia’s typical user experience, but they illustrate the ways user interpretation can go awry. At 3 billion bases in length, the genome is massive, and due to the paired, double-stranded nature of DNA, each position is associated with two nucleotides. Even getting oriented within this realm is a struggle, and SNPedia’s customers often get tripped up in determining whether the variant they have is good or bad.

Once you’ve identified what variant you have, decoding the literature is itself a complex task. Different studies of the same genes invariably find different effects. One study might report that a certain variant confers a 200 percent increased risk for some disorder; another finds only a 10 percent increase. Another still finds no association at all. The kinds of studies that form the bulk of SNPedia’s database are especially variable because of the technical complexity inherent in associating genetic variants with diseases.

In academia, a team of scientists might spend several years of effort to reconcile all of the various studies. Such an effort (called a meta-analysis) weighs the methods of each study, the number of the participants, and the statistical analysis behind every result to come up with a single best estimate of the effect of a given genetic variant.

SNPedia obviously lacks the technical expertise (not to mention time) to achieve the same level of rigor. Instead, SNPedia offers a subjective, poorly defined “magnitude” score, which corresponds roughly to how worried you should be about possessing a given variant. Cariaso’s idea is that over time, as more studies become available, the magnitude score will become roughly calibrated to the degree to which a given variant increases your risk.

Explaining the notion of a genetic risk factor is no easy task. Users have been primed by misleading media coverage to expect a single genetic variant to indicate whether they possess a certain disease. There are multiple stories in the forums of people attempting to self-diagnose or medicate psychological illnesses based on their genes. Since most variants typically confer only a slightly elevated risk of developing a particular disease, it is practically impossible to determine a medical condition based on one or even a handful of variants.

These misinterpretations and self-diagnoses can have potentially dangerous effects. For example, one Promethease user, citing information gleaned from a Promethease health report, refused to take the full dose of a doctor’s prescription. The same user blamed the medication for a series of kidney problems and internal bleeding the user later experienced.

For all of the confusion SNPedia generates, its efforts seem largely well-intentioned. Cariaso, who agreed that misinterpretation of genetic health information is a serious problem, patrols SNPedia’s forums and attempts to disabuse users of the most common misconceptions. Other sites are more predatory, playing on the scientific sheen of genetic testing to recommend vitamins, supplements, and alternative medicine.

On this end of the spectrum sits NutraHacker. For the price of $37, NutraHacker will produce a custom report that analyzes your genome and recommends (but does not sell) various vitamins, supplements, and foods. NutraHacker claims that its reports can help customers avoid harmful supplements, optimize their athletic training, and “detoxify.”

The trouble is that there’s little scientific backing for such a report. Most people in the developed world do not suffer from vitamin deficiencies, and when they do, it’s because of malnourishment or starvation, not their genes. While nutritionists hope that genetic information may in the future enable more precise vitamin prescriptions, the consensus is that the science isn’t advanced enough to guide recommendations yet. Nor could I find a single example of a genetic variant that has been linked specifically via vitamin consumption to improved athletic performance.

When I pressed NutraHacker’s founders, Mark Wagoner and Jon Vinea, on the issue, they produced dozens of articles on genes involved in vitamin metabolism, none of which directly supported their claims. Genetic variants typically exercise only modest effects on vitamin uptake, making them a poor basis on which to prescribe supplements, especially when most healthy adults already receive the full recommended dose of nutrients from their diets.

To get a feel for NutraHacker’s recommendations, I generated a report based on my own genome. Most of NutraHacker’s suggested supplements are innocuous: substances such as whey that have few serious health effects, so long as they are not taken to excess. In one case, however, NutraHacker “encouraged” me to take progesterone, a potent hormone used in female birth control pills, hormone replacement therapy, and over-the-counter menopause treatments. Since progesterone carries significant side effects and has the potential to cross-react with other commonly used drugs, this recommendation is reckless. And while NutraHacker is quick to encourage supplements, the reports explicitly refuse to specify dosage recommendations, opening the door for users to potentially consume supplements to excess.

NutraHacker isn’t alone in giving bad advice. Other services such as LiveWello explicitly advertise the idea that they can use your genome to tell how you will respond to various medicines. Without a doctor’s expert advice, consumers can misinterpret such advice with dangerous effects.

Ironically, in the FDA’s decision to prevent 23andMe from supplying health information, the agency expressed a desire to prevent consumers from misunderstanding their test results and “self-managing” their care based on genetic tests. But the FDA hasn’t actually stopped users from facing these dangers. It stopped sites that sequence DNA from also providing interpretations of the DNA, but it hasn’t stopped companies that supply interpretation only. Instead, customers have turned to a spectrum of less regulated companies that provide confusing and in some cases fallacious information.

In the FDA’s desire to protect citizens from themselves, it severely underestimated the lengths to which people will go—and the sources to which they will turn—in their efforts to understand their own genomes.