Avey proposes a new genomics paradigm.
By Kevin Davies
June 10, 2008 | BOSTON—While the celebrity quotient in the audience for the final Bio-IT World Expo keynote might not have matched some of the events Linda Avey has attended in recent months, she didn’t seem perturbed. “I feel like you guys are my peeps!” said Avey, co-founder along with Anne Wojcicki (wife of Google’s Sergey Brin) of personal genomics phenomenon 23andMe. “This is the group I grew up with.”
Avey reviewed 23andMe’s web-based, direct-to-consumer genome service and the company’s broader mission. 23andMe is about “how can we learn more about each other and about ourselves and build upon this information for yourself.” For $999, 23andMe takes an individual’s DNA sample and offers information on ancestry and genetic predispositions. A person’s maternal lineage can be traced via mitochondrial haplotypes from the maternal lineage, and paternal ancestry via the Y chromosome. Said Avey: “When you go to a bar, the pickup line could change from, ‘What’s your sign?’ to ‘What’s your haplogroup?’
Avey credited ex-Stanford geneticist Joanna Mountain with doing “a wonderful job of building out all the information in these maps.” Avey’s own mitochondrial haplogroup is H3, which she shares with about 70 “connections” on the site.
A new feature at 23andMe.com is ancestry painting. Avey showed a genome representation of an anonymous African American male, with chromosome segments colored navy for European ancestry, green for African. The donor was “64% European and only 33% African. One could maybe argue that this is what Barack Obama looks like.” Added Avey, “I would have shown you mine, but it’s really boring—all solid blue.”
The 23andMe “Gene Journal” provides personalized genetic information on various conditions that can impact a person’s health. Initially containing just 14 topics, the list of traits has grown quickly to 60 with new entries added monthly.
Like other personal genomics companies, 23andMe applies a vetting process to validate published gene associations. “People like to paint 23andMe as the more fun and frivolous company, but we very seriously take the science of what are the genetic discoveries coming out of the research community,” Avey said.
Avey cited a 2007 gene association on “avoidance of errors”, which although published in the prestigious Science magazine, only merited 1 star (out of four) by 23andMe’s in-house experts. “We feel like it’s almost the responsibility of 23andMe to put the caveats around this study,” said Avey. “The study size was probably very small, and it just didn’t hit all the criteria we’ve developed,” she said. When she hears customers dismissing an association because it only received a 1-star rating, she says, “that’s exactly what we want.”
For type 2 diabetes (T2D), 23andMe currently screens nine genes to assess a person’s relative risk of the disease (which has a major environmental component), and graphically displays the risk adjustment contributed by each gene. These risks can change depending on the individual’s ethnicity. “23andMe really wants to add more diversity to the studies that are being done,” said Avey. “Unfortunately, a lot of the work is done in Caucasian populations… We really want to try to change that, and make this applicable to all populations.”
Despite the flurry of recent genome associations, Avey cautioned: “We think it’s too early for this information that we’re sharing with people to take it in to a point where they’re doing too many things that are actionable.” Said Avey, “It’s really, really early for us to be making these big leaps of faith into what are the clinical endpoints for this information.” Many more genes will be uncovered for T2D and other common conditions, she predicted.
23andMe sees huge opportunities in social networking, not just for individuals wishing to share information a la Facebook, MySpace, and YouTube, but, “We’re interested in taking this phenomenon and moving it more into the research space. We see this happening in Health 2.0,” citing new resources such as PatientsLikeMe and Sermo that nurture networks for patients and doctors alike.
“We’re calling the next phenomenon Research 2.0,” said Avey. She foresees cancer survivors and adverse event patients sharing information online, looking for people with similar experiences, perhaps displaying badges on their Facebook pages when they contribute to future medical publications. Imagine, “if we can take these data back to the pharmaceutical industry, and say, ‘Here’s what we’re seeing, what are you guys going to do about this?’”
Through all this, Avey said 23andMe intends to retain the perspective of the consumer. “We’re really representing them. We’re here as caretakers of their data. The data belongs to them, it doesn’t belong to us.” Part of that goal is to launch a research project called “23 Diseases—areas we intend to do research in. We’ll create cohorts, we’ll bring people in.” The greater mission, said Avey, could be called, “23andWe. Let us know what you’d like to study and we’ll get started!”
The IT in Genetics
Many members of the 23andMe engineering team hail from Ebay, My Space, and the like. For a company that needed everything to happen “in web time,” traditional IT and database infrastructures weren’t going to work. “We have to show the whole genome,” said Avey. “We have to compare one person to many people, all in the time it takes to load a web page.”
Rather than put each SNP in a cell in a database, “They put all of our SNPs basically in a bag,” Avey said. In addition to traditional MySQL relational databases, 23andMe also uses “a proprietary file system-based genotype storage and custom indexing.”
Security is another key concern. “All of the data we have [are] encrypted… even the encryption keys are encrypted, so there’s been almost a paranoia within the company.” The main genotype data are stored separately from the web analytics. “We’re very sensitive that these are real people, and we keep this information very secure and private, even within our own science team.”
23andMe’s web services use a LAMP stack (Linux, Apache, MySQL and Perl/Python), which Avey liked in part because “it’s mostly all free.”
View the webcast of this keynote.
This article appeared in Bio-IT World Magazine.
Subscriptions are free for qualifying individuals. Apply Today.