In 1994, Mary Sue Kelly, who lives on a boat near Half Moon Bay, California, was diagnosed with breast cancer. Some 14 years later, she’s now well past the crucial 10-year survival mark. The treatment—mastectomy followed by standard chemotherapy—was successful. But for many years, Kelly, now 63 years old, remained unnerved. It was not so much by the implications that her contracting the disease would have for her own health, but rather by what her oncologist told her. “This has implications for your family,” he said—an ominous reference to possible heritable factors lurking in Kelly’s genome. “It just hit me in the stomach,” she recalls. “I was prepared to die, but my children—it just crushed me to think it could happen to one of them.”

So when a comprehensive test for global genetic-risk factors became available a few months ago, she did what a growing number of patients curious about their disease potential are doing these days: She took the test, offered by a just-launched company named Navigenics. The test came back with some reassuring news: She was actually below the national average for known genetic factors predisposing women to breast cancer.

“Folks who live in boats are not, by and large, people of the future,” Kelly says. But she’s fascinated by the new genomics. “Wouldn’t you want to know your blood pressure? Wouldn’t you want to know you have four cavities? I believe it’s the thing of the future. I suspect my grandchildren will have DNA information on their driver’s licenses.”

It looks as if the future Kelly foresees will soon be upon us. Companies are now offering whole-genome scans—probing entire collections of our genes for minuscule person-to-person differences at particular locations along the vast stretches of DNA that comprise our genes. These differences distinguish us from one another, tell us about our ancestral heritage, and influence our susceptibility to disease. At least one consumer-genomics company, Navigenics, offers follow-on genetic counseling services to explain to users what their results mean. And some critics are voicing doubts as to whether anybody really knows what they mean in these early days of genomic science.

As the consumer-genomics industry self-assembles before our eyes, the Federal government is doing its best to keep up. In April, Congress enacted The Genetic Information Nondiscrimination Act, a law protecting individuals against discrimination by insurance companies based on genetic test results, encouraging consumers’ use of such tests. “It has cleared the way for consumers to have access to their genetic information while safeguarding it from abuse, or misuse,” says Sara Katsanis, genetics policy analyst for the Genomics and Public Policy Center (GPPC), a Washington, D.C.-based Johns Hopkins affiliate. “This should also empower the companies themselves.”

What remains to be seen is how quickly medical practice geared for diagnosing uncommon, binary “you either have it or you don’t” single-gene disorders will be able to evolve to handle the rapid changes in store for it by dint of massive increases in the supply of, and the demand for, complex and ambiguous genetic data. The test Kelly took cost $2,500. Similar global tests (without some of Navigenics’ high-end follow-up) are on the market for $1,000. Researchers are pressing ahead at an accelerating clip with their efforts to identify and characterize the tiny blips in our DNA that can spell the difference between life and death—or between an aversion to broccoli and a hankering for it.

The fundamental building block of the edifice that is genomics is the so-called “genome-wide association study” (or GWAS, as it is called in the scientific community). GWAS (pronounced “gee-wahce”) and another acronym, SNP (“snip”), may sound as exotic to our ears today as “stem cell” did 10 years ago, but before very long both terms will be as familiar to us as “software” or “modem” are today.

These days, we all know that our genetic information is stored as long strings of substances called DNA base pairs. Effectively the genome’s smallest possible accounting units, base pairs—the pennies, as it were, of genetic variation—come in only four varieties. Strictly put, there is no such thing as “the human genome.” There are, rather, upwards of 6 billion individual examples. The reason we’re not all identical twins is largely because of the presence of minuscule “single nucleotide polymorphisms”—SNPs, for short—which are inter-individual variations in resident base pairs at particular points along the genome.

To fully appreciate the immense challenge of cost-effectively identifying SNPs, imagine your genome as two stacks (one for each parent’s contribution) of 3 billion pennies apiece on a table, all indistinguishable except for the fact that each penny bears one of four mint marks, analogous to DNA base pairs. Now, carefully align your maternal and paternal stacks side by side, and compare mint marks at each position counting up from the tabletop. (Don’t try this at home. The twin towers will be 1,700 miles high and cost $60 million.) Maybe once every 300 pennies or so, those mint marks differ: There’s your SNP. Where along the genome and with what frequencies these variations pop up is unpredictable—they have to be located through large numbers of genomics comparisons. To a significant extent, that’s been done.

Your pattern of SNPs tells you something about your ancestry. SNPs arise via rare germ-line mutations, spread through inheritance, and disappear only if the lineage bearing them dies out or if another mutation occurs at the same position. As a result, SNPs accumulate rather slowly in the genome over evolutionary time, so different populations inherit different sets of them.

Because a single base-pair switcheroo can affect the function or quantity of a protein coded by the gene in question, SNPs can help explain your little brother’s cowlick—although admittedly the environment (he never did comb his hair) probably played a larger role. In some so-called “high-penetrance” conditions such as sickle-cell anemia, cystic fibrosis, or hemophilia, a single SNP in a single gene can cause disease.

Those monogenic disorders are often dire enough to preclude successful mating, so the total number of patients who have them is relatively small. The common chronic diseases of adulthood that account for the bulk of all sickness and death in the developed world (for example, cardiovascular disease, cancer, and late-onset diabetes) are “multifactorial”: not the result of one funky gene variation, but rather the product of a complex interplay of several SNPs, each contributing a small increase in risk, with the environmental stage on which our genetic scripts play out.

That’s where the genome-wide association study (the GWAS mentioned earlier) comes in: Take a large group of people with a disease (“cases”), and another equally large bunch without it (“controls”), and scan their genomes for as many SNPs as your technology and funding permit. If a given SNP pops up with significantly different frequencies between cases and controls, it may be a genetic risk marker.

Tatiana Foroud, who as head of Indiana University School of Medicine’s genomics division has conducted numerous GWASs, cautions that cases and controls must be well matched for age, gender, and ancestry, or you’ll get spurious results. The presence, for example, of many people of a particular ancestral type in the case group, and few or none among the controls, may cause SNPs associated with that ancestral lineage—as opposed to ones that really have a causal role in disease—to be inappropriately flagged. Foroud also says the bigger the GWAS—more subjects, more SNP locations monitored—the more robust the data.

Until fairly recently, the idea of snooping for SNPs on a genome-wide basis and ferreting out those responsible for small increments of risk to a patient’s profile was simply not feasible. But high-resolution SNP-detection microarray technologies (notably platforms sold by Illumina and Affymetrix) have made possible the accurate genotyping of as many as 550,000 or 1.8 million locations, respectively, at a reasonable cost. This, in turn, has hugely enhanced investigators’ ability to comb genomes for combinations of elusive, low-penetrance genetic variations.

Meanwhile, an offshoot of the Human Genome Project called the HapMap Project has radically reduced the number of SNPs that actually need to be examined. It would be prohibitively pricey to monitor all 10 million estimated SNPs lurking in the human genome (about 6 million have actually been identified), but fortunately you don’t have to. The HapMap Project showed that SNPs come in clusters—if two of them are close enough together, an offspring inheriting one will almost always inherit the other, too, so genotyping one is enough. Furthermore, many of them turn up in parts of the DNA that don’t code for gene products or their regulation, at least as far as is now known.

The new SNP arrays and the HapMap Project have fueled a wave of GWASs that are hitting the medical literature on a weekly basis. SNPs associated with late-onset diabetes, cardiovascular and neurodegenerative diseases, a variety of cancers, and more have been isolated and, in varying degrees, characterized. Taking advantage of this flood of new genetic information, several ventures have geared up to mass-market whole-genome scans. Two Internet-oriented startups, both physically located in California’s Silicon Valley—23andme.com (Mountain View) and Navigenics.com (Redwood Shores)—give users access to a Web-based report within two weeks after they spit into a petite vial shipped to them when they sign up.

23andme, which began offering its service to users last November, charges $1,000 and studiously avoids attributing medical utility to its findings, emphasizing instead SNP findings that “tell users interesting things” about their ancestry or about their normal traits—things like aversion to or inability to taste a particular vegetable component. “We’re not a healthcare company,” says co-founder and co-president Anne Wojcicki. “We don’t know exactly how some of this genetic information should be used in the clinical setting.”

However, 23andme does provide estimates of relative risk for about 10 disorders and has also augmented the 550,000-SNP Illumina platform it uses with a customized panel of some 30,000 SNPs associated with individual differences in drug metabolism, excretion, and toxicity. (The company hasn’t started reporting those results to users yet.)

In contrast, Navigenics, which launched in April on an Affymetrix platform that genotypes almost 1 million SNPs, is explicitly assigning medical significance to the results its users get. “All disease has a genetic component, and if you ignore it you’re not making the best medical decisions,” says Navigenics cofounder and Chief Scientific Officer Dietrich Stephan, a trained geneticist. Navigenics’ $2,500 service includes follow-on telephone consultations with genetic counselors. One hour is reserved for each consultation, and users aren’t limited to just one session. They can call anytime—before taking the test, after taking it, in the middle of the night—if they have concerns, at no extra cost.

Elissa Levin, the board-certified director of Navigenics’ genetic counseling program, says, “We take their family and personal medical histories, and try to explain: What does this mean to them and to their family members? What are the next steps for them to take?”

Navigenics has used rigorous criteria to select the 18 indications it reports on to customers, Stephan asserts. “They’re the common causes of morbidity and mortality, the genetic risk factors for them are well fleshed out in peer-reviewed studies, and they’re actionable—if you learn you’re at increased risk you can do something about it. There are known environmental risks to avoid, or early-screening strategies to take advantage of.”

Levin says it seems a lot of people take Navigenics’ test because they’ve had or are worried about getting one of the diseases it predicts, or because they know someone else who’s had it. For these customers, sometimes no news is good news. On the other hand, discovering a totally unexpected risk factor can be a spur to action.

“People usually think of me as being on the thin side for my age,” says 53-year-old Terry Dorbert, another Navigenics user. Curious about her missing family history (she’d been adopted), Dorbert was surprised to find that she had a higher than average genetic risk for obesity and late-onset diabetes. Although Dorbert says her own behavior hasn’t been altered by her results—she already exercises a lot and eats carefully—her college-attending daughter, with whom she has shared her results, has since dropped 20 pounds.

Some critics argue that any claim of medical significance to SNP variation patterns is premature, as the prospective case/control studies required to clinically validate their predictions have yet to be done. “Unlike a pill that goes on the market, these tests have not been subjected to any sort of third-party review, let alone FDA approval,” says the GPPC’s Katsanis. “If you predict that a population has a 20 percent higher risk of diabetes, you have to follow them up and find out if they get it.” That will take years, she says. And it will cost somebody a lot of money.

Navigenics’ riposte is that early detection of a predisposition to a late-onset disease can pay off in increased vigilance, behavior change, and in some cases early medical intervention.

Levin recalls a 40-year-old participant in Navigenics’ early beta-testing who learned she was at significantly increased risk for colon cancer. Medical guidelines recommend colonoscopies beginning at age 50, and few people are compliant even at that age. But the woman’s physician authorized one at her request. A 1.5-centimeter polyp was found and removed. “If she had waited 10 years or more, that most likely would have developed into cancer,” Levin says.

“We’re 100 percent confident that the markers we pick are real, validated genetic risk factors,” says Stephan, who adds that Navigenics imputes medical significance to only a fraction of the close to 1 million SNPs their platform records. An in-house team of statistical geneticists vet all GWAS papers reported in the medical literature, he says, with only a handful meeting the company’s criteria: typically, large, well-designed studies, independently replicated and published in top peer-reviewed journals. 23andme makes a similar claim.

Both 23andme and Navigenics are preparing to incorporate outcome information from consenting users into the burgeoning anonymously coded genomic databases they maintain, and both expect to eventually collaborate with outside academic researchers on follow-on studies to clinically validate the SNP findings shaping their reports. Executives of both outfits emphasize that they are taking great care not to release data without consent or in any form that could compromise customer privacy.

Navigenics is collaborating with the Mayo Clinic on a study now underway to see whether learning about their genomic risks changes users’ behavior. But what’s certain is that consumer genomics is going to change the behavior of the medical community. Sequencing your entire genome isn’t down to pennies yet—at the moment it retails for $350,000. But nobody doubts that the price will drop to more like $1,000 within five years. Cheaper, more widespread sequencing means that increasingly rare but sometimes significant SNPs stand to be unearthed, leaving a growing mountain of genetic data tapping its foot to be interpreted.

Is the medical community ready to deal with the inevitable torrent of partially digested genomic-scan data that will soon be begging for interpretation? Virginia Thurston, Foroud’s colleague at Indiana University, recently published a study suggesting otherwise. Thurston, the medical school’s medical-genetics program course director, suspects schools are producing primary physicians who may lack basic skills such as taking family histories or explaining implications for patients’ families, and who may be fuzzy on essential principles of genetics.

In all, there are only between 3,000 and 4,000 total board-certified genetics specialists in the United States, according to Thurston—a shortfall unlikely to be relieved anytime soon, because the field isn’t lucrative. A dermatologist sees three or four patients per hour and performs lots of procedures. A genetics specialist spends an hour to an hour and a half talking to patients and their families—hardly the formula for sizeable reimbursements.

Plus, virtually all practicing specialists’ experience and training is restricted to single-gene traits. Few of them have expertise in the complexities of low-penetrance gene variants churned up by genome scans, says Jonathan Holt, a medical geneticist and Stanford University clinical instructor. Holt recently decided to start his own company, Holistic Genomics, to counsel “walk-ins”—spillovers from genetic testing by 23andme and other consumer-genomics outfits, such as Iceland-based DeCodeMe, that proclaim their genome scans are “for information purposes only” and offer no genetic-counseling follow-up.

For their part, consumer-genomics companies say they are working hard to train the medical community by, for example, providing good online materials for users and more-detailed printouts they can bring to primary physicians.

Navigenics has funded the construction by an independent provider, Medscape, of a continuing medical education course for primary-care physicians about genomics and what it could mean for their practice. “We thought it was important to have a third party develop it, to ensure that it would be for the pure benefit of physicians,” says Levin. The online course was first offered in late February. By mid-April, more than 2,000 physicians had already completed it.

It’s still early days for consumer genomics. “We are very clear that our test is definitely not for everybody,” says 23andme’s Wojcicki. “People who aren’t interested in preventive medicine, who don’t want to know their genetic information, shouldn’t get tested. It’s a personal choice. But people should have that choice.”

Navigenics user Perry Pickert, 32, accepts that his genomic-scan findings, which showed him at below-average genetic risk for the colon cancer his father had recovered from and the prostate cancer his grandfather had died of, are subject to alteration by new findings. “This isn’t a Bible,” Pickert says. “I’m not closing my eyes.”

Bruce Goldman is a San Francisco-based science writer and futurist whose work has been published in The Journal of Life Sciences, Nature, Science, New Scientist, The Los Angeles Times, The Journal of the American Cancer Institute, and, most recently, Nature Reports Stem Cells. He is the co-author of two books about how technology is shaping our destiny.