Here Is a Human Being (7 page)

George saw money less as something that would taint and commoditize personal genomics than as a means of altering the subject-researcher power dynamic. Indeed, he suggested he might be willing to blur the scientist-subject distinction altogether. While some bioethicists have argued that returning genetic research results to subjects is a good idea, it’s not clear how many feel that way. (Some have issued “consensus” statements,
⁵⁴
but as far as I can tell, while putting out consensus statements is a favorite pastime for bioethicists, it tends to correlate poorly with actual consensus.) For their part, all the data say that participants would very much like their results back.
⁵⁵

Certainly there are good reasons
not
to give experimental results back: research is just research, after all—it has not been clinically validated. Do the researchers even know what the data mean, especially if they’re not clinicians? And what if the information they give is wrong—can they be counted on to clean up the fallout from results that mislead or, more likely, just don’t pan out?

George didn’t buy it: he had long thought disclosure of genetic and genomic results to subjects was the natural thing to do. He said he would like to make PGP subjects true partners in the outcome of his project. Yes, they would bear the risk of eating from the tree of genomic knowledge,

however bitter and uncertain that might taste, but they would also reap the rewards. “We’re hoping everyone in the study will not only know what’s going on, they will actually be working with us to analyze their own data. You could be so informed you might even qualify as a coauthor.”
⁵⁶

He said that returning results to subjects was imperative because it would give them the information they needed to make the single most important decision about the PGP that other human genetics protocols did not: when to quit. “I think the opt-out clause of most consent forms is a mirage,” he told me. “It’s a fake. If you don’t get your information back, then how would you know you have a Huntington’s mutation and don’t want that to be in the public domain or even a private database? Ideally, opting out means that that type of information would be erased from everything. But if you don’t know it to begin with, then you can’t ask the investigators to erase it.”
⁵⁷

“So what will you do with your exome?” I asked him in early 2007.

“Um … what do you mean? I’m going to
study
it.”

“Okay. But will
I
be able to study it?” I felt like a six-year-old boy at summer camp: I’ll show you mine if you show me yours.

He demurred and suddenly sounded more like the modest, private man Ting described to me. “I’m not going to be superfast in putting mine in the public domain. I intend to, but I think I would like to be a guinea pig for the phasing process where I’ll look at it myself as much as I can with software, and then get some experts involved who are inside the PGP and look at it with them with moderate security. And then together if the PGP subjects and the PGP researchers feel that there’s anything that needs to be redacted, then we will redact it either from the genotype or the phenotype or both. We will leave behind a scar that says ‘this was redacted.’ We won’t say, ‘Oh no, George doesn’t have schizophrenia.’ We will say, ‘We’re not
saying
whether George has schizophrenia or not.’ And those scars will be revealing in a certain sense, but they may not be actionable. We’ll work with ELSI scholars and genetic counselors and genetic experts to try to proactively redact, and not just for me but for everybody else in the PGP who wants to do this.”
⁵⁸

George told me his daughter wanted her personal genome—or some approximation of it, anyway—for her sixteenth birthday; he and Ting had tentatively agreed. I met her at the first PGP barbecue at the Church-Wu home. She was a tall, striking teenager with her mother’s jet-black hair and her father’s liquid, penetrating eyes. And while she didn’t like school (“not my thing”), she seemed to share her parents’ propensity for creativity and overachievement. She is an artist: her abstract photos, paintings, and multimedia creations adorned the walls of the house. She told me the plate I was eating off of was one of hers. She designs clothes. And she’s a fashion model (“I’m trying to sign with a big agency”).

She shared three short-term goals. “On my birthday I’m gonna go to school and tell them I’m not coming anymore, I’m gonna get a tattoo, and I’m gonna get my genome done.” Why did she want her genome? “I’d really like to know what’s coming. If I’m gonna have a short life or if there will be uncomfortable things in my future, I really wanna live now.” This was the flip side of Jim Watson: he needed
not
to know at the end of his life, she needed
to
know at the beginning of hers. They seemed to share a fatalistic, deterministic view of their genomes.

I asked if she thought she was unique or whether her friends were thinking about their personal genomes, too. She assured me they were, but qualified her answer like a true scientist. “This is Brookline,” she said. “Harvard is right down the street. So this isn’t a random sample of teenagers. A lot of my friends know about it because of my dad and also because their own parents are aware of it.”

“How much of your and your friends’ interest in personal genomics is related to MySpace and Facebook—people just putting everything on the Internet?”

“I really don’t think any of it has to do with that,” she said. “I just think we’re a more informed generation because of the Internet in general. Anyway, MySpace is out now—it’s all old men and creepy people.”
⁵⁹

“Playing in bad rock bands,” I offered.

“Pretty much.”

George said that five thousand dollars for a set of genotypes (the projected price for a scan of a half million genetic markers in mid-2007) was not much of a lifetime investment to make on behalf of a teenager. And pretty soon, everyone would be doing it.

Even before the recent explosion of personal genomics, there were signs that the ossified, neglected, backwater specialty of medical genetics was about to change. Jason Bobe started the blog The Personal Genome: Genomics as a Medical Tool and Lifestyle Choice in 2003.
⁶⁰
In early 2007, just before taking a job helping George to manage the PGP, he told me he was planning to launch a new website, GenomeHacker.com, designed to give young people crude ways of interrogating their genomes without the benefit of fancy lab equipment. He expected they would be able to infer things about their DNA simply by studying their own phenotypes; for example, if they couldn’t drink coffee late in the day without being up all night, it was likely they were slow metabolizers of caffeine and therefore harbored a genetic variant in one of the major genes that encodes a drug-metabolizing enzyme. “Ten years from now there’s going to be a whole bunch of fourteen-year-olds developing tools for parsing their genomes,” said Bobe. “It’ll be a new after-school hobby for kids, I imagine.”
⁶¹

Bobe, a curious and precocious guy who turned thirty in 2009 and has become a friend, was enthusiastic about technology and life as depicted in the pages of
Wired.
He liked to send emails at 4
a.m.
; his Gmail status message often revealed how many unanswered emails were currently in his inbox (two hundred to four hundred seemed to be the norm). He was stocky and still spoke with a midwestern twang; he credited his blog with rescuing him from the Indiana cornfields.

I never doubted the viability of his idea: genome hacking as an after-school endeavor. But it had already become clear that it was not going to take ten years. Or five. Or even one. Neither kids nor their parents would have to infer their genotypes from their phenotypes. They could go right to the source.

W
hen I met Matt Crenson on an overcast Wednesday morning, his deep, raspy voice belied an affable manner, which was appropriate: his employer, the Mountain View, California–based company 23andMe, had for the last several months taken great pains to portray itself as friendly and nonthreatening—the
anti-Gattaca.
George Church, who was on the company’s scientific advisory board (he was on some eighteen other such boards, too; God knows how he found the time), often described 23andMe as “playful.” The lobby of the company’s home in a nondescript Silicon Valley office building looked something like a hipster toy store that was building up its inventory before officially opening its doors for business.

The atmosphere—a basal bustle of twenty-somethings, occasional peals of laughter echoing among the cubes and glass offices—harked back to the dot-com era: there were shelves filled with Day-Glo-colored squishy rubber balls while the otherwise Spartan lobby was scattered about with cute stickers, buttons, and plastic packages of Mike & Ike’s red and green candies festooned with the company logo and its catchphrase, “Genetics Just Got Personal.” (I stuffed several in my bag for the plane ride home.) 23andMe thought of itself not as a health-care company or as a biotech, but rather as an Internet start-up. “Web 2.0” was the descriptive phrase one heard over and over from its employees.

23andMe officially launched in November 2007, just a couple of days after the first commercial entrant into the personal genomics market, deCODEme, which was an outgrowth of Icelandic genetics-meets-pharma firm deCODE Genetics. Both companies began by offering customers access to what I will call “the variome.” If your genome is all 6 billion DNA base pairs (the function of most of which we don’t understand), and your exome is the 20,000+ genes (about 60 million base pairs) that code for protein, then your variome is a smaller subset still: it is an assortment of markers more or less evenly spaced across the genome that tend to vary from person to person; some markers fall within genes, but most do not. By early 2010 researchers had identified nearly 13 million of these markers; from 2007 to 2009 the companies typically typed 500,000 to 1 million of them (about 1–2 million base pairs total). These marker sets (called single nucleotide polymorphisms, or SNPs—"snips”) were thought to capture much of the variation in human DNA, although they represented no more than 0.05 percent of the entire genome.
^*
But they were relatively cheap to type given that such endeavors would have cost hundreds of thousands of dollars in the 1990s; both 23andMe and deCODEme began by charging customers a thousand dollars. Customers entered their credit card number online, waited for their kits, spit in a tube, put it in the mail, and a few weeks later could log onto a secure website and look at their variomes.

Crenson used to be a science reporter for the Associated Press. Given the downward trajectory of the newspaper business, he seemed genuinely happy to land on his feet at 23andMe as “Content Manager,” the guy responsible for editing everything the customer read. This particular morning, however, the customer account he tried to show me, belonging to the whimsically named “Greg Mendel,”
^*
wasn’t working. After some tinkering and technical help from a young guy dressed in black, we finally got to meet Greg Mendel’s genome. Matt walked me through it and pointed out various risk alleles for prostate cancer. “This one raises his risk to 1.29, but this one has only been studied in African Americans. There’s a lot of uncertainty now about how much ethnicity affects the results.”
^†

Crenson explained the company’s criteria for including a trait among the company’s officially sanctioned list, what it called the “Gene Journal.” To make the Gene Journal, a trait had to have been studied in a thousand people or more, been independently replicated in another study, and been reported in a reputable, peer-reviewed journal. “Not the
Albanian Journal of Medical Genetics and Dentistry,”
Crenson assured me.
¹
In December 2007 the company featured eighteen traits in the Gene Journal, from earwax consistency to restless legs syndrome to Crohn’s disease. Within a few months there would be a total of seventy-eight traits. Soon it was no longer called the Gene Journal, but simply “Health and Traits.”
²
By 2010 the company was reporting risk estimates on forty-seven “clinical” traits that it considered to be fully vetted (including twenty-one recessive diseases for which one might be a carrier), and eighty-seven additional “research” traits for which either there was insufficient data (in 23and-Me’s eyes) or the traits did not affect disease risk (for example, “food preference,” “hair color”).
³

From the beginning 23andMe offered customers information about their ancestry using mitochondrial DNA (passed on only by mothers) and Y-chromosome markers (passed on only by fathers); eventually it began including markers from the other twenty-three chromosomes (1 through 22 plus X). Crenson took me back in time through “Mendel’s” genealogy and the groups of markers that had been transmitted down through his family. A few months after my first visit, 23andMe would incorporate social networking based on genomic characteristics into its menu: a Facebook for the genome-savvy set. At the Global Economic Forum in Davos, Switzerland, company executives passed out spit kits to the glitterati, an apparent strategy to help brand personal genomics as hip.
⁴
Bono was tested. Jimmy Buffett and Warren Buffett were tested: no relation, they learned!
⁵
Peter Gabriel was tested.
⁶
In 2009, 23andMe launched an online community of pregnant “mommy bloggers”: “Explore the genetic legacy your child will inherit from you and your partner.”
⁷

23andMe was cofounded by Anne Wojcicki, the wife of Google zillionaire Sergey Brin, and Linda Avey, formerly of biotech behemoth Affymetrix and its now-defunct human genomics spinoff, Perlegen. On the day I visited, true to 23andMe’s start-up mojo, both women were running around with great urgency, putting out fires and hurrying to and from meetings and conference calls. They apologized and asked if I could come back later. That afternoon, Avey, a striking blonde in her late forties who grew up in South Dakota, finally sat down to talk about the origins of the company.

When Avey was at Perlegen, her mission was to convince pharmaceutical companies to use the Affymetrix GeneChip technology—the small glass wafers used to type some of the millions of SNPs in human and other genomes—to begin to go after genes that cause specific diseases. But in those days the genome was terra that was even more incognita than it is now—relatively few SNPs had been characterized. Thus Perlegen couldn’t begin gene discovery without first embarking on an expedition to isolate more markers that would allow it to sharpen the cartography of the genome. That meant a $100 million commitment to sequence bits and pieces of fifty human genomes, however crudely.
⁸
This was an arduous slog, but it paid off: the company managed to identify 1.7 million SNPs, an unprecedented treasure trove in 2003.
⁹
(For comparison, when I began graduate school in the early 1990s we knew about no more than a few hundred polymorphic DNA markers across the entire human genome.)

Perlegen was now in a position to design a DNA chip with several hundred thousand markers and begin to do genome-wide association studies (GWAS). These are essentially very dense case-control studies designed to find DNA markers important in disease. By typing the same set of markers in large numbers of cases and controls, it becomes a brute-force statistical matter of finding markers that pop up more frequently in sick people than in healthy ones. Those markers are very likely to be in or near genes that play a role in disease.
¹⁰

GWAS studies have since become ho-hum.
^*
But only a few years ago, they were the new new thing. So new, in fact, that NIH and industry were reluctant to fund them. By 2005, Perlegen had designed a chip with 250,000 markers, identified cohorts with diseases such as Parkinson’s, but could get neither corporate backing nor public funding to move ahead with GWAS.
¹¹

“You might identify a group of two thousand samples,” Avey said, “but the NIH funding mechanisms were stuck in their old paradigm—they had cutoffs. People would look at their sample sets and say, ‘Okay, [given my funding] I can only afford to do a certain number.’ And then it would sort of defeat the purpose because you wouldn’t have the statistical power to really find what you were looking for. The whole vicious cycle was infuriating after a while. A lot of people at Perlegen and Affy were frustrated. This technology was just sitting there and being under-utilized.”
¹²

It was this frustration, Avey told me, that planted the seed for a freestanding personal genomics company. “What if,” she thought, “we just shifted this paradigm and opened up the ability for people to pay their own way and gave them access to their own genetic information?”
¹³

Her first idea was spas. Spas, she reasoned, had become more clinically oriented while doctors’ offices had become increasingly more spalike. “You had this weird convergence happening. People at spas have got disposable income and they are very interested in their health. Some of the high-end spas had changed their image to become ‘wellness centers.'”
¹⁴

Avey took the concept to some of the Affy VPs with whom she went on six-mile runs in the morning. “Talk to Steve,” they told her. According to Avey, Affy CEO Steve Fodor was enthusiastic and told Avey she had to do it … but not at Affy. It was, he told her, beyond the firm’s core mission as a research tool provider.

At a meeting where Perlegen was presenting data on the company’s efforts in Parkinson’s disease, Avey met Google mogul Sergey Brin, whose mother suffered from Parkinson’s.
¹⁵
Brin began asking Avey detailed questions about Perlegen and its analytic approach. Hopeful that she could persuade Google to back her new venture, Avey began trying to set up meetings with Brin, but found it difficult.

“He always wanted to have his girlfriend there with him. And I thought, you know, who is this girlfriend?”
¹⁶

Anne Wojcicki had been investing in health care and was herself frustrated at the sector’s lack of progress and poor return on investment. She and Avey came together at the annual Technology, Entertainment and Design meeting in Monterey and agreed to move ahead with what they called—aptly if unimaginatively—"Newco.” Google would kick in $3.9 million and Newco would eventually become 23andMe. Wojcicki told
Fast Company
that she was in her kitchen reading Wikipedia, saw a picture of the twenty-three pairs of chromosomes, and started singing the words, “Twenty-three and me.” Newco had a new name.
¹⁷

Where 23andMe went for playful, its Silicon Valley neighbor, Navigenics, opted for serious. Serious as a heart attack, one might say. Whereas the 2009 iteration of the 23andMe Web presence was all light and shiny and iPod nano–ish, the Navigenics website felt more like a doctor’s office, full of pastel and sepia tones, handsome and mature couples, all sun-swept and looking extremely healthy—one half expected a medical history questionnaire and six-month-old magazines to materialize on one’s screen. The original video I saw describing the company’s Health Compass service was presented by cofounder and chief scientific officer Dietrich Stephan, a short, amiable, and balding man in his early forties. In the video he appeared to me to be somewhat uncomfortable and constricted; in person he reminded me of a gentler, more garrulous, and perhaps more rumpled version of the great B-movie actor John Saxon. When I first met him at company headquarters he wore baggy pants and layers—shirt, red fleece, and brown cotton jacket. He spoke softly and with a faint Pittsburgh accent; he laughed easily, even when talking about the minutiae of FDA regulations.

He began plodding along from square one, giving me a rudimentary history of the Human Genome Project and Craig Venter. I tried not to make my impatience too obvious but rather to politely move him beyond the scripted VC speech. When I asked him about the impetus for starting Navigenics, he became animated, just as Linda Avey did when talking about the birth of 23andMe. Stephan was, he told me, toiling away at a large genomics research shop in Arizona, the Translational Genomics Research Institute (TGen), finding disease susceptibility genes and watching his discoveries go … nowhere. “I started getting frustrated that people weren’t applying this to understanding individual risk profiles. I tried desperately to get things placed into molecular diagnostic facilities, but they didn’t understand how to interpret that information and give it back to physicians or genetic counselors. They would say, ‘This is not a binary diagnostic.’ That’s what all of these folks are used to: you have two copies of a cystic fibrosis mutation and you’re going to have cystic fibrosis. Or you have one copy and you’re a carrier. Or you have zero copies [and you’re neither].”
¹⁸

The diseases Navigenics focused on initially—type 2 diabetes, Crohn’s disease, heart disease, multiple sclerosis, obesity, rheumatoid arthritis, and a dozen others—rarely played by Mendel’s simple rules of genetics, that is, one gene = one trait. One can carry half a dozen versions (risk alleles) from six different genes that predispose to a common disease and still be at average or even below average risk for developing that disease. Stephan immediately recognized the challenges Navigenics would face. “How do you put that information in context and how do you communicate risk to people?”
¹⁹

When describing the field, he used the word
nuanced
several times. He spoke about educating physicians and genetic counselors and developing a gold standard. “We want to set the bar high. We’re going to come at this from a hard-core medicine and science perspective, put all of the spokes of this wheel in place and then roll it out.”
²⁰