November 4, 2021
Kristen V. Brown
A few months ago, on the morning 23andMe Holding Co. was about to go public, Chief Executive Officer Anne Wojcicki received a framed sheet of paper she hadn’t seen in 15 years. As she was preparing to ring in the Nasdaq bell remotely from the courtyard of her company’s Silicon Valley headquarters, Patrick Chung, one of its earliest investors, presented her with the pitch document she’d shown him when she was first asking for money, reproduced on two pieces of paper so she could see both sides. The one-sheet outlined a radical transformation in the field of DNA testing.
Wojcicki’s plan back then was to turn genetics from the rarefied work of high-end labs into mainstream health and quasi entertainment products. First she’d sell tastemakers on her mail-in spit kits as a way to learn sort-of-interesting things about their DNA makeup, such as its likely ancestral origins and the chance it would lead to certain health conditions. Eventually she’d be able to lower prices enough to make the kits broadly accessible, allowing 23andMe to build a database big enough to identify new links between diseases and particular genes. Later, this research would fuel the creation of drugs the company could tailor to different genetic profiles. 23andMe would become a new kind of health-care business, sitting somewhere between a Big Pharma lab, a Big Tech company, and a trusted neighborhood doctor.
Some of this still sounds as far off now as it did during the Bush years. Improbably, though, 23andMe has rounded second base and is heading for third. Wojcicki did sell millions of people on DNA test kits—11 million and counting—and bring such tests to the mainstream, with some help from Oprah’s holiday gift guide. An estimated 1 in 5 Americans have turned over their genetic material to 23andMe or one of its competitors. Now that she’s got the data, Wojcicki is working on the drugs. Her company is collaborating on clinical trials for one compound (and nearing trials for another) that could be used for what’s known as immuno-oncology, treatments that attempt to harness the body’s complex immune system to beat cancer. 23andMe says it’s also exploring drugs with potential use in treatments for neurological, cardiovascular, and other conditions, though it declined to specify them. Last month the company bought Lemonaid Health, a telehealth and drug delivery startup that offers treatment and prescriptions for a select group of conditions, including depression, anxiety, and STDs.
Chung, now a co-founder and general partner at the venture capital firm Xfund, says his framed gift to Wojcicki was meant to emphasize how exceptionally closely she stuck to her original vision for 23andMe over the past decade and a half. Most young tech companies in search of their initial venture funding have little hope of identifying their eventual path to great wealth and instead focus their energies on some sort of “dancing startup laser light show,” he says. “The rule, not the exception, is that whatever the founders are pitching at the very earliest stages is simply never the thing that is successful.” 23andMe went public in June at a market value of $3.5 billion, is now closer to $5 billion, and is expected to report about $56 million in revenue in its latest quarter, though it’s unlikely to turn a profit.
Of course, Wojcicki, a former Wall Street research analyst with a focus on biotech, did have some experience with outstanding startups. Soon after she closed that round of venture funding, she married longtime boyfriend Sergey Brin, who co-founded Google in her sister Susan’s garage. (Wojcicki and Brin divorced in 2015.) It’s easy to see the Google influence in 23andMe’s strategy—collect all the data, derive whatever insights you can, and find an adjacent line of business with the potential to yield much bigger profits. Even today, Wojcicki’s conversations tend to be shot through with an older strain of Web 2.0 techno-optimism about ways to better connect people with, in this case, how their medicines get made, and to cut through the randomness and waste that suffuses the science of drugmaking. “One of our core values is, like, we’re all in this together,” she says.
“One thing I always think is a tragedy is that you develop a drug and then people hate you. I’m really interested in, can we actually be the first drug development group that is loved by people?”
The next phase of her master plan might sorely test that question. While it’s difficult to imagine anyone saying they love their pharmaceutical company, it wouldn’t be crazy for the 8.8 million 23andMe customers who once absently checked a box saying yeah, sure, use my data for whatever, to feel like they’ve been bait-and-switched now that their genes are laying the groundwork for potential cancer cures.
Privacy advocates have been warning for years that the spit-tube deal is lopsided—that there aren’t enough legal protections on genetic data to justify trading DNA samples for answers about whether you’re predisposed to hate cilantro or what percent Swedish you are. DNA data, which contains information about you and your blood relatives, could be hacked, de-anonymized, or shared with the police. 23andMe’s pharma ambitions add a new dimension to these concerns. If Wojcicki keeps achieving her goals, customers might one day pay 23andMe to assess their disease risk and pay for a treatment it later develops based on their DNA. Why should one company hold the key to the world’s genetic code—and charge the rest of us handsomely for access to it?
And before 23andMe reaches that threshold, it will have to deal with the reality that Big Pharma is hard. The company is trying to expand from a category of modest consumer products into a field full of goliaths a couple of orders of magnitude larger, with research and development operations that dwarf anything it’s fielding. “Genetics is just one little piece of the puzzle,” says Harvard geneticist Robert Green. “So far there have only been a few instances where genetics has been the key to a brand-new drug. It’s not yet clear it’s going to have the value we’re hoping it does.”
For now, Wojcicki’s answer to both sets of concerns is, basically, Trust Us. “I always like to say, ‘We’ll win you over,’ ” she says.
From the moment the first draft of the human genome was published in 2001, scientists have evangelized the potential of genetic data to discover and develop drugs. That’s because Big Pharma’s process, for all the billions of dollars in annual R&D spending, remains a crapshoot. The mysteries of human biology mean that, more often than not, promising drug candidates, even those that make it deep into clinical trials, wind up not working. The industry calls that period between the lab and the market the Valley of Death. An estimated 13% of all global drug development programs end in approvals. Even then, drugs can inexplicably work well for some people and not others. To this day, researchers sometimes can’t explain the molecular mechanisms that underpin common medications.
Given enough data, genetics can identify patterns that provide significant clues. When you’re hunting for what might be a few letters of variation among 3 billion pairs of DNA nucleotides, scale is what matters. Often, many genes contribute to a particular condition to varying degrees, so more examples give researchers more chances to puzzle out how they fit together.
Wojcicki and her co-founders weren’t the first to think of building a database for this purpose, but using consumer DNA kits as an input soon gave her company a massive scalar advantage over pretty much everyone. (The only bigger known databases belong to Ancestry.com and the Chinese government.) 23andMe started selling its kits in 2007, pitching them mostly as a way to assess a smattering of health traits, such as a customer’s risk of colorectal cancer or likelihood of lactose intolerance, along with a rudimentary version of its ancestry analysis. Criticism soon followed. One article, published in the journal BMJ in 2008, argued that “rather than improving health, widespread genetic testing is likely to result in widespread anxiety.” On these sorts of grounds, the U.S. Food and Drug Administration forced the company to stop selling its tests without approval in 2013. Other critics questioned whether consumers could self-report their own health data accurately. 23andMe relies on customers to help its analysis along by flagging known health issues—whether a given person has a heart condition, for example, or an autoimmune disorder.
Eventually, with copious peer-reviewed research, 23andMe won over the scientific community and the FDA, which allowed it to resume selling the tests in a pared-down form in 2015. Today the company sells an ancestry test for $99, one with health “insights” for $199, and a subscription service that includes reports on heart health and a person’s particular response to different medications. For most Silicon Valley companies, hoarding the test-kit data and selling it to drugmakers would have been the endgame. Instead, 23andMe is trying to make the drugs, too.
The company’s 100 staff scientists, hired over the past six years, work out of its therapeutics offices in South San Francisco, half an hour north of its headquarters in Sunnyvale, Calif. While the Sunnyvale campus is all glass and contemporary opulence, the therapeutics lab is a drab, squat complex with a 1980s flavor. (On the other hand, it’s tough to knock the sweeping bay views.) The scientists worked through the pandemic here, at lab benches spaced for social distancing, with thick cables running from the benchtops to the ceiling.
The facility is divided roughly into halves, each with its own offices and labs, by a big room for shipping and receiving and an even bigger set of storage chambers for gases, chemicals, and distilled water. On one side, robots make the drug prototypes. There are robots for pipetting DNA into cells, robots for purifying liquids, and even robots that just swirl containers of liquid for a certain amount of time. On the other side, scientists suss out the basics of how genes and diseases are connected, often performing tasks by hand at their benches.
Aside from the research scientists, the staff is mostly working remotely, and the rows of empty cubicles lend the more corporate parts of the office the feeling of a startup in the aftermath of the dot-com bust. The labs, though, have remained busy. During the company’s last fiscal year, from April 2020 to March 2021, it roughly doubled its number of good potential drug candidates, as evaluated by its scientists, to 19. Their focus: diseases that 23andMe has lots of data on, that many people suffer from, and that lack good treatment options. “We prioritize based on power, need, and speed,” says Kenneth Hillan, head of the therapeutics division.
Every six months or so—after, say, a big post-holidays influx of customers—23andMe’s computational biologists scan the company database looking for variations that, based on correlations in the company’s massive database of human DNA, appear more common among people with particular diseases. Turning that hint of where there might be treasure into a map where X marks the spot takes many months of research in the dreary building full of scientists. But the database has the potential to narrow to months a process that otherwise might take many years.
While pharma companies typically specialize in a few kinds of treatments, 23andMe can instead direct its research wherever the data takes it, and it has a ready supply of possible future clinical-trial participants and customers. The end result is the potential to develop drugs more quickly and efficiently than the status quo. This has made it an attractive partner to traditional drugmakers, notably GlaxoSmithKline Plc, which took a $300 million stake in 23andMe in 2018. Within a year, GSK and 23andMe already had six potential targets in the works. One candidate, a cancer drug that works to block CD96, a protein that helps modulate the body’s immune responses, entered clinical trials in four years. The industry average is closer to seven.
In the case of the drug 23andMe is developing in-house, scientists discovered a biological pathway, or series of molecular actions that lead to a change in a cell, strongly associated with the way immune cells attack a certain kind of tumor. The drug is designed to block the tumor-induced suppression of the immune system’s T cells and then reactivate the immune system’s response to the tumor.
Sometimes the database can also highlight new uses for existing drugs. “The size has opened up the opportunity to be able to see genetic associations that you can’t see anywhere else because you need the math of the numbers to be able to see it,” says John Lepore, GSK’s head of research. At one point, a genome-wide association study showed that multiple genes within a particular pathway appeared to be linked to asthma. There was already an antibody drug in development for asthma by another company that works by blocking the action of that pathway. 23andMe scientists were able to further interrogate the database to find that drugging that pathway might also help to treat a certain rare disease.
Wojcicki is now gearing up to shift more toward solo projects. That was a big reason why 23andMe went public in June, via a merger with VG Acquisition Corp., a special purpose acquisition company, or SPAC, founded by billionaire Richard Branson. By March of next year it expects to have its first drug developed fully in-house in clinical trials, and it says its Phase I trial with GSK should return data by the end of 2022.
Skeptics, however, argue it’s unrealistic to expect the company to be great at drug development right from the start, especially as it decouples its efforts from Glaxo and other experienced developers. “If they try to travel this road by themselves, it will be perilous,” says Bernard Munos, a researcher and senior fellow who focuses on biomedical innovation at the Milken Institute, a conservative think tank. “It’s like sending a man to the moon. On paper it may look fairly straightforward, but the details are what will cause the enterprise to succeed or fail.”
Before the company’s second post-SPAC earnings report, on Nov. 10, Wojcicki says she’s still working out what the integration of Lemonaid will look like. For starters, 23andMe will train Lemonaid’s fleet of doctors on how to make use of genetic information and try to keep its services cheap enough that insurance doesn’t have to get involved. (Lemonaid’s services, like consumer DNA kits, aren’t covered.) “When I think back on the original promise of genetics, it’s been adopted in cancer treatment but not at all in primary care,” she says, adding that the company is trying to figure out what actionable health recommendations based on genetic risk factors should be. In the early going, that will most likely mean using its pharmacogenetics reports—which help determine which drugs are likely to work best for an individual—when prescribing drugs through Lemonaid doctors. Your genes can affect how well you metabolize certain drugs and may influence, for example, which antidepressant or antifungal works best with your biology.
Eventually, Wojcicki says, that might lead 23andMe to be involved in customers’ health care in two ways: using the consumer testing to head off illness while the drugmaking side manages those that can’t be prevented. “In five years,” she says, “what I am looking to be measured on is, can we keep people healthier?”
As for whether these new business lines might exacerbate customers’ privacy concerns, Wojcicki’s main counterargument is that the U.S. medical establishment is lousy on privacy, too. For years the buying and selling of health data has expanded dramatically without much doctors or patients can do about it, if they’re even aware. Unlike when you go to the doctor, she says, 23andMe customers have a choice about how their data are used.
Well, kind of. There are no federal laws prohibiting companies outside of a health-care setting from providing individuals’ genetic information to third parties, and the existing protections of genetic data in the U.S. are weak at best. That became clear in 2018, when police used a different, open source database called GEDmatch to make an arrest in the long-cold Golden State Killer murders. Suddenly consumers everywhere were very aware of just how serious the consequences of sharing your DNA can be, which apparently made them less enthusiastic about home DNA kits. 23andMe’s sales dropped off, and layoffs followed in early 2020. While calls to strengthen consumer DNA protections died down during the pandemic, 23andMe’s latest development may help to reignite those efforts.
“They’re transparent, but only to a certain degree,” says Jennifer King, a privacy and data policy fellow at Stanford’s Center for Internet and Society. “My data could be extremely valuable to them.” King says a better system would require a third party to broker data and make sure consumers are compensated fairly.
In some cases, after all, one individual can hold the key to a world of biomedicine. Take the famous case of Henrietta Lacks, whose family struggled in poverty for years after researchers turned her cancer cells into a critical research tool that made millions of dollars. With a far greater range of the human genome decoded, it’s easy to envision a Gattaca-esque future in which the DNA of the masses is mined for personalized miracle cures affordable only to the super rich.
Wojcicki says that’s just not going to happen. “We’re not evil,” she says. “Our brand is being direct-to-consumer and affordable.” For the time being she’s focused on the long, painful process of drug development. She’d like to think she’s earned some trust, but she hasn’t come this far on faith.