Skip to main content
Premium Trial:

Request an Annual Quote

As Whole-Genome Sequencing Advances, Optimists Find Empowering Knowledge; Others See Dizzying Data


By Turna Ray

The optimism behind genomically guided personalized medicine was palpable at a recent gathering of life science pioneers who are among the first people to have their genomes sequenced. The guests at the Genomes, Environments, Traits Conference — held in a glass-paneled room overlooking the Charles River in Microsoft's research and development center in Cambridge, Mass. — were a self-admittedly exclusive bunch that are testing out the waters ahead of the sea-change that individual genomic data is expected to bring to healthcare.

For these early adopters of personal genome sequencing, there are several reasons to be excited: The cost of whole-genome sequencing is dropping rapidly toward the $1,000 mark; new business models around personalized healthcare are emerging; and every day a fresh crop of studies are published, providing more insight into the genetic underpinnings of disease and response to treatments.

Although whole-genome sequencing is currently available to a select few, the hope of early adopters is for this technology to be accessible to all. At the meeting in Cambridge, attendees projected a timeframe of between five to 10 years for when genome sequencing would be more readily available to the general public.

Yet some proponents of genomic medicine question this timeline. Muin Khoury, director of the Office of Public Health Genomics at the Centers for Disease Control and Prevention, told Pharmacogenomics Reporter that while he would love to see whole-sequencing become available to the masses, the genomics landscape is currently littered with data that is unvalidated, untranslated, and of questionable utility. The challenge of sifting through noise to identify markers of use makes it difficult for Khoury to share in the optimism of early adopters.

One early adopter, Illumina CEO Jay Flatley, estimated at the GET Conference that the industry is about a decade away from being able to offer sequencing to everyone. Flatley was the first person to get sequenced through the company's personal sequencing service when it launched last year (PGx Reporter 06/11/09). Factoring in reagent costs, the service is currently priced at $48,000.

"People should get sequenced to [contribute to] finding solutions to diseases and make people's [health] better," urged Life Technologies CEO Greg Lucier, who also has had his genome sequenced. Life Technologies contributed to lowering the cost of sequencing earlier this year by launching the Applied Biosystems SOLiD 4 Sequencing System, which the company said can sequence a whole human genome for $6,000 per genome.

As leaders in the nascent field of personalized medicine — who have beyond a basic understanding of genomics and who are perhaps more comfortable dealing with probabilistic estimates of disease risk than the average person — the GET pioneers seemed outwardly unaffected by the concerns commonly raised by the less optimistic when the topic of genomic testing is raised. At the conference, fears such as genetic insurance discrimination and social stigmatization resulting from genetic determinism were dismissed as unlikely. Political, monetary, and ethical barriers were not dwelled on. And the more immediate hurdles to adoption, such as analytical challenges and education, were acknowledged, but not discussed in detail.

"I wasn't worried about the privacy thing," said Esther Dyson, who was the third person to be sequenced through George Church's Personal Genome Project. Initiated in 2006, the Harvard Medical School-hosted PGP has sequenced and published the exomes of 10 people so far and is aiming to do whole-genome sequencing for 100 volunteers. Ultimately, the PGP wants to enroll 100,000 informed participants form the general public.

Flatley similarly added: "I don't think the insurance company is going to analyze my genome any time soon"

And the always outspoken James Watson, co-discoverer of the structure of DNA, maintained that broader application of the technology would be a public good, waving off the "ethical stuff" as "just crap."

Countering the celebratory mood of the GET Conference, that same week a paper appeared in The Lancet analyzing the clinical value of Stanford University Bioengineering Professor Stephen Quake's whole genome sequence. In annotating one man's genome, a feat which took nearly two years, researchers illustrated some of the challenges in mainstreaming whole-genome sequencing that may have gone unaddressed at the conference.

In the study, 31 researchers from Stanford University and other institutions analyzed 2.6 million SNPs and 752 copy number variations in Quake's genome sequence to identify three rare variants associated with cardiac sudden death, several mutations in CYP2C19 suggesting a positive response or lack of response to various drugs, and numerous other variants associated with disease risk and drug response.

Quake by no means represents the average consumer, having sequenced his own genome with the Helicos platform, which he invented. He quipped at the GET conference that "it was very easy to get informed consent" for the clinical analysis of his genome.

However, in an accompanying opinion piece in The Lancet, the authors of the Quake annotation paper point out that even if whole-genome sequencing is available and properly interpreted, educating the public to enable informed decision-making from the data "will be difficult, lengthy, and expensive."

[ pagebreak ]

Although genomics has long promised to make healthcare more efficient and reduce costs, rolling out genetic testing in a broader healthcare setting could actually increase costs, the authors noted. "For example, many variants identified through whole-genome sequencing will be benign, but sequencing might result in a large increase in testing by cautious physicians to rule out false-positive results," they wrote. "Such an increase would not only raise healthcare costs, but would subject patients to the physical and psychological costs of increased testing."

Annotation of Quake's genome also presented a clear divide between the clinical utility of genomically guided disease risk stratification and pharmacogenomically guided treatment decisions. Quantifying genetic predisposition to diseases and analyzing pharmacogenomic responses "are two activities that are extremely different," said Russ Altman, a professor of bioengineering, genetics, and medicine at Stanford University and one of the researchers who analyzed Quake's genome.

This division "came out in the analysis because the folks doing the disease risk were very nervous and they didn't know how to interpret lots of things. They kept saying, 'If we tell someone they're at risk for certain things, and what if we're wrong. What's the data?'" Altman told Pharmacogenomics Reporter last week. "Whereas for the drugs, we had the exact same uncertainties, but sometimes the data was very good and so we were just confident that it was right. And other times we said we can just tell the doctor to be careful."

Given the level of manual curation necessary to characterize Quake's disease risks and the uncertainties that still remain in the analysis, Altman predicted that mainstream application of whole-genome sequence analysis to assess disease predisposition was still two decades away. In contrast, Altman believes that PGx-guided drug administration strategies can be broadly rolled out in physicians' offices in a five-year timeframe.

"Ten years would be a disappointment for the drug-related stuff," Altman added by e-mail. "I'm hoping five years. I'm an optimist."

When CDC's Khoury heard the five-year timeline, he had a very different reaction. Khoury said that the utility of genomics in healthcare is currently "a moving target," and that it is "too early and premature to put these things in practice right now."

While Khoury is in favor of whole-genome sequencing becoming broadly available as a long-term goal, as a public health professional must also shoulder the burden of assessing what's possible in the near term given finite resources.

The PGP-10 "are the people who have drunk the Kool-Aid, obviously, and the people who are the early adopters," said Khoury, who was not among GET Conference attendees. "We still haven't figured out what to do with genome-wide association studies. And GWAS is only [dealing with] 1 million genetic variants. So, what to do with 3 million base pairs just escapes me right now."

Urging for more large-scale, randomized studies to translate how genomic data impacts health outcomes, Khoury warned that annotating one person's genome sequence suggests little about the broad applicability of this kind of approach. "If you do this on a 100 or 1,000 people, it can help three people and hurt 15 people. And then what have we really done here?"

Unraveling the Molecularly Open Man

Harvard's Church put it best at the GET Conference when he observed that while early adopters reap the advantages of a new technology before anyone else, they also "take a lot of the body blows" for being the first.

In discussing their data and experiences openly, these genome sequencing guinea pigs are essentially allowing the public to dissect their decisions, actions, and reactions. They may be optimistic, but these early adopters are also trying to grapple with what exactly the data means, what it could mean, and what it cannot reveal.

While the social and ethical barriers to broadly introducing whole-genome sequencing may not have been discussed in great detail at the meeting, sequencing pioneers are nonetheless challenged by them.

Nearly all the GET speakers shared a complex process by which they arrived at the decision to have their genome sequenced. For the PGP-10, this involved passing an exam that tested their understanding of genomics and ethical issues. For others, it involved months of self-education and lengthy discussions with family members and reluctant doctors.

Even though some claimed that they aren't worried about privacy issues or discrimination, they have nevertheless proceeded cautiously with regard to how their sequence data might impact family members, and have at times blinded themselves to certain information that they themselves would rather not know.

[ pagebreak ]

This process of challenging one's perceptions of "personal information," is exactly what Church said he intended when he launched the PGP, which sequences volunteers' genomes for free with the condition that they publish their data. While this exercise is likely not an easy one, it can reveal the often contradictory and arbitrary nature of the information that many consider private, and the data that most people readily post on Facebook.

Watson joked at the conference that Church should encourage more volunteers for the PGP, which has a lengthy informed consent process, by "giving them a pie or something." However, in Watson's case, his age was the sweet spot for him deciding to get his whole genome sequenced. "If I were 20 I wouldn't want to know because you just worry," he said. "At 80, I don't worry if I'm going to get cancer."

It is also well known by now that the genomics pioneer didn't want to know everything his genome held. Watson had his APOE status redacted from the results due to a history of Alzheimer's disease in his family.

Other pioneers at the conference expressed concerns about what their genome sequence data might mean for other family members. For example, after Church published his genome data through the PGP, his wife Chao-ting Wu admitted she was concerned about how people would receive her husband's genome sequence data and she was particularly worried about whether people would make any negative inferences from Church's data about their daughter's genome. "By having George's genome sequenced, I was concerned that [people] would know half of [our daughter]," Wu said at the conference, adding that her husband did not pressure anyone else in the family to get sequenced.

Similarly, Illumina's Flatley said that his decision to sequence his genome was made after discussing it with his family and his doctor, who initially refused to write the prescription necessary for Flatley to go ahead with testing. Unlike most direct-to-consumer genomics firms, who maintain that doctors should not be intermediaries between consumers and their personal genomics data, Illumina requires a prescription from the customer's personal physician (PGx Reporter 06/10/09). Flatley's wife has decided against having her genome sequenced, he added.

John West, CEO of Novocell and a former executive at Solexa and Illumina, said that after getting 23andMe's personal genome scan, he and his family decided together last year to also be sequenced through Illumina's sequencing service. The ethics of the West family's decision to be sequenced "for non-medical purposes" has been publicly questioned, particularly with the involvement of two teenage children. However, it is rarely noted that West suffered a pulmonary embolism five years ago, and subsequent genetic testing revealed he had a genetic mutation for the inherited blood clotting disorder Factor V Leiden. "There are risks if you haven't been sequenced," West said. Whole-genome sequencing has been "medically actionable for our family."

Many of the pioneers said that the fears associated with getting sequenced and making that data public are just symptomatic of the early stage of the field. Some likened the concerns of social stigmatization or insurance discrimination associated with genome sequencing today to when X-rays were first introduced and people felt the technology was a privacy intrusion. Others compared consumer apprehension with genomic analysis to the early days of the internet when people were reluctant to enter their credit card information online. However, conventional wisdom among early adopters holds that these apprehensions will dissipate as whole-genome sequencing becomes more integrated within healthcare, particularly among the more uninhibited "Facebook generation."

At least one member of the Facebook generation, Anne West, 17, isn't ready to unveil her genome profile to her online social network buddies. Although her father has submitted his sequence data to GenBank, West said at the GET Conference that she will consider publishing her sequence data when she is 21, but she'll likely wait until she is much older to actually do so.

Church has always acknowledged that the decision to be genetically tested or sequenced is not a decision that should be entered into lightly. In response to a high school student's question during the conference about whether she should be genetically tested in light of her mother's reservations, Church said, "If you and your family have doubts then you should wait."

Roll Out the Good News

From the stories of sequencing pioneers and, more acutely, the clinical evaluation of Stephen Quake's genome, it is clear that the public may be more ready to embrace genomic testing in the context of drug response, as opposed to disease risk.

"Once patients see their [doctor] using genetics to optimize their drug dose, they will ask, 'Why the heck don't you do this for all my meds?' and then we will have very, very fast uptake and penetration," Altman said via e-mail. "Not so much for the disease risk stuff, which is mostly bad news.

"Telling someone they are going to get diabetes, that's a big deal," Altman said. "But telling them that they may not respond that well to a drug [just means] the doctor can use a different drug."

The drug and insurance industries have embraced pharmacogenomics, particularly in the area of cancer. Large pharmaceutical companies are increasingly inking partnerships with diagnostics firms early in the development of drugs to ensure that they'll be able to identify patients most likely to benefit and least likely to see adverse reactions. This can be good news for patients, as they will be able to avoid unnecessary toxicities, and for drug companies, since they will be able to differentiate their genetically targeted products from non-differentiated competitors in the marketplace.

The oft-cited barrier to genomic testing adoption, physician education, may also be less daunting of a challenge in the context of pharmacogenomics, in Altman's view. "When making treatment decisions, doctors already account for their patients' medical history, behavior, and family history, and so on," Altman said. They are quite used to making prescribing decisions with "multiple sources of fuzzy data," he pointed out. "This is another reason I think [genetic information in the context of] drugs are very different.

"You don't have to train [doctors] in genetics," Altman added. "You just have to say that the genetics indicates with high confidence or medium confidence or low confidence that this drug may cause a higher incidence of side effects."

[ pagebreak ]

The value of genetic testing could also be presented in a context that is familiar to doctors. Physicians today "are trained generally in pharmacology, as part of medical school," Altman noted. "So, when they do ask, 'Why?' you can tell them that this drug is going to have different pharmacokinetics, that it's going to stay in the bloodstream longer, or it's going to be eliminated faster. It could really be a one liner and it really builds on training that they already have had and experience they already do have in making drug choices."

Altman feels that getting doctors comfortable with genetic testing in the PGx context is a manageable goal for the next five years, and he is working toward that end through his efforts to create a genomics certification program for physicians. Altman is on the accreditation board of the Scripps Translational Science Institute's newly formed Association for Genomic Medicine. With a $300,000 grant from Life Technologies, Scripps plans to launch by year-end an online academy to "try to credential as many physicians as possible in genomic medicine," according to Eric Topol, director of the Scripps Translational Science Institute (PGx Reporter 05/07/10).

Keeping Physicians in Mind

With regard to the broader applicability of genome sequencing to assess disease predisposition, the researchers analyzing Quake's genome also kept the physician closely in mind. In order to identify three rare variants — TMEM43, DSP, and MYBPC3 — associated with sudden cardiac death, and a variant in LPA in line with a family history of coronary artery disease, researchers applied an analytical method that hasn't previously been used in genomics, but one that doctors are much more used to.

Through painstaking manual curation of published literature and lengthy searches through healthcare databases, researchers estimated Quake's post-test probability for getting a disease by multiplying his probability for getting the disease without accounting for genomic data with likelihood ratios for every allele associated with increased risk for that particular disease. A patient's pre-test probability factors in things such as social environment, geographic residence, ethnicity, and age, while post-test probability takes into all those things plus genomic risk.

Instead of analyzing pre-test and post-test probabilities, most DTC genomics firms, including 23andMe, Decode Genetics, and Navigenics, use allelic or genotypic odds ratios reported in published studies to provide lifetime risks for various diseases. Not only does this not give people the whole picture about the actual contribution of genetics to their predisposition for diseases, but doctors are much less attuned to dealing in odds ratios for their patients, according to Atul Butte, assistant professor of medical informatics and pediatrics at the Stanford University School of Medicine.

"At least 30 years of physicians have been taught to understand pre-test, post-test, and likelihood ratios. It's required knowledge to pass board exams in most sub-specialties today," said Butte, who was also an author of the recent Lancet paper on Quake's genome. "When we were in medical school we used to have carry a little slide rule with us that you could just slide it up and down to get the pre-test probability and the likelihood ratio and it would compute the post-test probability."

Butte worries that after getting a genome scan that conveys risk in terms of genotypic odds ratios, people might see they are at lower genetic risk for a particularly prevalent disease, such as prostate cancer, and stop getting annual screenings. Looking at it in terms of likelihood ratios, "That would be a terrible thing to do, because even if you have no elevation in genetic risk for prostate cancer, the pre-test probability [might be] so high that you're still more likely to get it," Butte said.

While the researchers have published the pre-test probabilities for Quake, which could be conceivably applied to someone of the same age, ethnicity, and background, they have not curated similar data for people who fall into other demographic categories. However, Butte is hoping to get proxy measurements from other countries that keep track of this type of data and payor groups in the US, which would allow for this data to be expanded.

"A number of insurance plans do keep track of those data, and eventually do make them available for research. And we're going to see if we can form collaborations with them to see if we can get better handles on pre-test probabilities," Butte said. "As we start to put electronic health systems in place in the US, I'm optimistic that we'll start to keep track of these numbers."

Data Downer

While the analysis of Quake's genome is a promising advancement toward broader applicability of whole genome sequencing, for some the barriers to adoption overshadow the promise. As proponents of DTC genomics and early adopters of genome sequencing often talk about the sense of empowerment they feel with access to their genomic data, public health officials like Khoury worry about whether that data will lead people to make decisions harmful to their health, like, for example, forego prostate cancer screenings.

"To think that more knowledge is power is fine, but I question the idea that more data is power," Khoury said. "What we have now are data points. The data are confusing at best. Some of it is knowledge, but most of it is not really knowledge."

[ pagebreak ]

In Khoury's view, before thinking about expanding access to sequencing technology, stakeholders should invest in validating genomic associations in large, randomized studies, and then make sure that their findings are clinically useful, which has been the mantra of many other researchers and health regulators.

"I don't mind if the PGP-10 had their genomes sequenced, but deploying this on a population-wide basis could wreak havoc on the healthcare system," Khoury said. "We're already in so much trouble in terms of unnecessary healthcare costs," he added, noting that improperly validated data could lead to increased testing and inaccurate healthcare decisions.

Khoury is leading an effort called GAPPNet, which is a CDC and National Cancer Institute-led effort to drive the adoption of validated and clinically useful genomic technologies (PGx Reporter 07-15-09). Its four key activities include: convening individuals and groups conducting genomics translational research, programs, and policy activities; sponsoring new translational research; synthesizing and evaluating available research findings; and developing and disseminating "validated, useful genomic knowledge and applications for use in medicine and public health."

The editorial accompanying analysis of Quake's genome articulated many of the data analysis challenges that could slow down broader application of whole genome sequencing. For example, annotation of Quake's genome took around 20 months, which involved manual curation of genetic disease and pharmacological risk. The authors pointed out that this information is hard to find and keep up to date, and there is no central repository for rare and disease-associated variants.

Furthermore, interpretation of the genome "involved the work of a clinical geneticist, a genetic counselor, and experts in bioinformatics, genetic cardiology, internal medicine, and pharmacogenomics, among others," the authors said. As such, they concluded that "efficient risk analysis of whole-genome sequence information will need urgent improvement of methods, with substantial automation."

Whole-genome sequencing itself has its limitations, too. The technology does not "reveal translocations, large duplications or deletions, copy number repeats, or expanding triplet repeats," the editorial pointed out. Additionally, the authors noted that whole-genome sequencing can't tell whether gene variations are in copies of genes on different chromosomes or in the same copy of the gene, which is a critical distinction necessary for recessive disorders.

Another thing that became clear from analysis of Quake's genome is that doing this for more people will require results to be delivered through the web. This raises access issues but also presents exciting new projects for NIH to fund in the areas of user interfaces and consumer education, Altman pointed out.

"Since not everybody in the world has great web access … this will probably have to be on cell phones, because one thing you can count on everyone around the world having, even very poor people, is cell phones," Altman said.

Illumina has said it is developing a prototype of a cell phone application that will allow people to see their whole-genome sequence data on their iPhones. Sequencing enthusiasts imagine a time when people can submit their blood samples to a lab directly through their phones and instantly view their results, but that's still in the realm of science fantasy. For the present, the technology is hobbled by more pedestrian concerns. "How can you, in a little tiny screen, tell somebody about this disease risk they have?" Altman posited. "This is a big problem."

Ultimately, even after all the analytical problems are solved and little phones are able to display the secrets of the human genome, where the science may ultimately fail is in actually changing human behavior.

Since Quake is "borderline" for three major risk factors for coronary artery disease, and has a prominent family history of vascular disease, his physician recommended statin therapy, noting he has gene variants that will allow him to respond well to the drugs. Quake said at the GET conference that while his genomic risk for coronary artery disease got him to the doctor's office, he has decided to hold off on taking statins for the time being.

Beyond the Quake project, other researchers looking at the impact of genomic information and human behavior have found the relationship complex. Much of the data on the topic suggests that genetics may incrementally alter people's actions related to their health, but on the whole patient behavior is hard to change (PGx Reporter 03/26/08).

When Altman writes a prescription, he often wonders if the patient is going to toss it in the trashcan on the way out of the office. "The number one reason for a drug not working is that the patient didn't take it," Altman said. Referring to Quake's data, he added, "so, that's a great bottom line, which is patients will still be patients. [Personal genomics data] will be a little bit of incremental information, but will it change compliance? That would be a miracle."