Ernest Hemingway’s writing may have tended to the short and sharp, but the man himself was apparently fond of the cuddly and extraneous, at least when it came to kittens with too many toes. A sea-captain friend of Hemingway’s, it seems, persuaded him to take in a polydactylic cat, and that cat became the progenitor of a colony of overly toed felines thriving today in and around the museum in Key West that was Hemingway’s home. The patterns of inheritance among those cats have even helped shed a bit of light on certain defects in human DNA. And so it is that Papa retroactively became an early contributor to the science of the human genome.
I learn this from Nadav Ahituv, a rising-star geneticist at the University of California, San Francisco, Medical Center, who studies the genetic roots of limb-related defects, obesity, and drug absorption. Congenital defects hold a special significance for Ahituv: As a teenager, he was in a body brace for three years with scoliosis. “It was not fun,” he says. “I spent my time reading Kafka, and then I started in on genetics textbooks, thinking that if I could understand what had gone wrong with me, maybe I could find a way to help others.” It must have been time well spent, because Ahituv, a fit-looking fellow who gets to work by propelling a folding bike up the city’s famously steep hills, has shown an uncanny knack for tracking human traits and disorders down to specific sections of DNA. He can now point out limb-altering segments of chromosomes as easily and precisely as he might map out coffee shops in the Lower Haight.
This is exactly the sort of progress we’ve come to expect from the triumphant cataloguing of the human genome, first accomplished in rough form in 2000 after a decade of work, and then polished up by 2003; researchers have since been amassing and analyzing genomic data at an ever-accelerating rate. The genome, as we all know, largely determines what we look like, our traits, and, significantly, our susceptibility to disease and other disorders. Ahituv is one of tens of thousands of well-funded researchers around the world trying to determine which segments of the genome contribute to which disorders. It is one of the biggest scientific endeavors in history, premised on the notion that the results can be used to prevent or fix many things, or possibly everything, that ails the human body — from allergies to cancer to aging itself. Dozens of biotech companies have sprung up in the past decade to commercialize this work, and one might assume that a stream of miracle pills will soon be on its way to our pharmacies.
You bet — just as soon as we work through a couple of hitches in this grand genomic enterprise. Scientists have indeed been superb at finding connections between disorders and various strips of DNA. But it turns out that in the vast majority of cases, these connections happen to be hideously convoluted, with any one disorder related to many genes and any one gene affecting many things in the body. Even when researchers are able to highlight a clear relationship between a single gene and a single disorder, they generally have little or no idea how those chunks of DNA are causing problems. Then there’s the disturbing tendency of gene-related treatments either to fail to work on the vast majority of people or else to entail horrific side effects. “Most individual elements of the genome can be perturbed, and there won’t be any substantial consequences,” says John Sninsky, vice president of discovery research for Celera in Alameda, California, the first company to sequence the genome. “Except for when there are catastrophic consequences.”
Yes, we’ve cracked the genome. Experts can identify every one of the 3 billion bases in every micrometer of DNA in any cell in your body. But so far, that has given the medical world no more ability to treat or predict most illness than knowing that Al Qaeda is camped out in Waziristan has allowed the U.S. government to clean up terrorism or predict where it will strike next. In fact, while thousands of links have been catalogued in journals and trumpeted in the media, with precious few exceptions virtually no promising new treatments or even highly useful diagnostics have emerged. And the situation is unlikely to improve much anytime soon. I happened to speak with prominent genetics researcher Russ Altman of Stanford on the 40th anniversary of the Apollo moon landing, and he was worried that the public might ultimately see genomic research in the same way: as an amazing feat with little practical payoff. “People said, ‘All that effort to put someone on the moon, why the heck did we do that?’ ” he told me. “Apollo did have lots of smaller spin-offs, like better jet engines, but I just hope the impact of the genome on our lives is more obvious than that 40 years from now.”
To understand the gulf between the genome’s promise and its payoff to date, it helps to understand why everyone was so excited about it in the first place. Conventional drug discovery basically involves semi-randomly throwing every known substance against every known disease — one at a time — in the hope that occasionally one of them will do more good than harm. Amazingly, that process has led to most of the world’s successful drugs and is still employed today. In the 1980s, scientists and biotech executives decided they could do better by analyzing the proteins that underlay a disorder — almost everything that happens in the body is mediated by proteins — and then custom-designing molecules to tie up the bad proteins or help out the useful ones. But proteins, complex molecules whose functions change depending on the even more complex ways in which they fold and the still more complex ways they interact with other proteins, turned out to be too hard to analyze in most cases. Then, as new technologies for identifying genes started to emerge in the late 1980s, scientists and the biotech industry were struck by a new vision: Why wrestle directly with proteins when you might track disease right to its prime source, the genes themselves?
For scientists, moving from proteins to genes felt like the difference between trying to reverse-engineer an alien spaceship and being handed a kit with color-coded tabs and easy-to-read instructions. The basic concept of the genome is by now pretty familiar. Inside each of our cells are 46 coiled-up, scraggly looking blobs — chromosomes — comprising long strands of the DNA molecule. These strands consist of interspersed pieces of your father’s and your mother’s genetic code (sperm and egg cells each have 23 chromosomes). The DNA varies along its length by the presence at any point of one of four different component molecules called “bases.” The patterns that these four different bases form along certain segments of the DNA determine the formation of specific proteins. A segment of DNA that specifies and helps produce a protein in this way is called a gene, and there are hundreds or even thousands of genes in each chromosome. Taken together, all the genes in the 46 chromosomes, along with the so-called junk (the segments of DNA that don’t “code” for proteins), are your genome, essentially a chemical blueprint for producing and operating a You.
And, the theory went, fixing You would be just as straightforward: Just note down the genetic code for both sick and healthy people, then look at where the code for sick people differed from the code for healthy people, and zero in on those strips of code — the genes of illness. Even better, the biotechnology community would soon figure out how to do it all on a chip! Now they’d have Moore’s Law on their side.
It cost $3 billion to sequence a human genome for the first time, accomplished with mostly public funding in a sort of distributed Manhattan Project parceled out to labs around the world. In the giddiness that accompanied that achievement, scientists, executives, and journalists talked up the coming fruits of this labor much as nuclear-power fans in the 1950s touted the advent of uranium-powered flying cars and robot maids. To take a few examples from The New York Times‘s relatively restrained coverage: “Drug discovery will rely as much on software and databases of genetic information as it will on test tubes and white mice”; “new cancer drugs … could send conventional radiation and chemotherapy the way of medicinal leeches”; heart bypass surgery would become obsolete when in “a glimpse of the real promise of genomics … a gene injected directly into a diseased area prompts the body to grow new blood vessels”; and genetic tests would provide each of us with a “kind of fate map” that would spell out disease vulnerability and even longevity. Meanwhile, investors and pharmaceutical giants began hurling money at biotech companies. Millennium Pharmaceuticals got more than $1 billion from a group of pharmas, and Human Genome Sciences took in $125 million from SmithKline Beecham. Both startups were based around genome-driven drug discovery. “People were raising money on fragments of genes, claiming they would lead to a new drug, a new cure,” says Frank Eeckman, an MD and biotech analyst for market-research firm NeuroInsights and a former genetic researcher on the original Human Genome Project at Lawrence Berkeley National Laboratory.
How has that all worked out? I stop by a biotech conference focused on emerging treatments for central-nervous-system disorders such as Alzheimer’s, Parkinson’s, stroke, epilepsy, depression, and schizophrenia — all targets of about 20% of all clinical drug trials going on now. Researchers have been talking up the promise of genome-related breakthroughs for these widespread and difficult-to-treat illnesses for years. But of the roughly 50 companies at the conference, not one is focused on approaches related to tracking down new genes. I ask former researcher Manuel López-Figueroa, a rock-star-looking vice president at prominent biotech VC firm Bay City Capital and a manager of a major academic research consortium, to tell me what genome-related treatments or tests are emerging in the field. He thinks for a minute. “As far as I know, nothing,” he says, finally. “People were very optimistic about DNA studies, but I can’t recall anything that has come out of them. Time will tell whether we’ll eventually get there or not, but would I put money into them? Philanthropic and government money, yes; investor money, no.”
I also look up all of the gene-focused companies mentioned in nine longish miracle-of-the-genome articles that ran in The New York Times and Boston Globe between 1998 and 2002: Of the 14 companies described as leading the way to remarkable new drugs and tests, all but one are out of business by virtue of having either folded, melted away in an acquisition, shifted to third-party gene-testing services, refocused on conventional drug development, or stooped to selling controversial direct-to-consumer products. The one exception is Human Genome Sciences, which traded as low as 45 cents a share in the past year, though its stock recently climbed dramatically after the company reported positive results from a midstage clinical trial of a lupus drug.
In truth, scientists knew something was wrong as soon as they counted up the genes after cracking the first genome. Until then, there were thought to be about 100,000 genes, and it seemed pretty damned remarkable that an entity as complex as a human being could be specified by so few pieces of code. But when the genes in the finished genome were tallied, they came to about 25,000. “At first, a lot of people had the hubris to think, Oh good, this will be even easier than we thought. We’ll just stick all the gene code in an Excel spreadsheet and work with them there,” says Celera’s Sninsky. But wiser researchers heard alarm bells. If there were one-fourth as many genes, then each gene was on average doing four times as much as they’d thought. How? The answer wasn’t long in coming.
It turns out that many dozens or even hundreds of genes each contribute to any given human attribute, and any one gene might contribute to several. Genes, in other words, turn out to work not as simple disease switches, but in impossibly complex networks. To see how big a setback this was, consider the hunt for so-called oncogenes — the genes of cancer — a search that has long grabbed a significant share of the $6 billion sunk annually into cancer research in the United States. As researchers ran endless gene studies — they’re still running them today — the number of genes seemingly involved in various forms of cancer steadily grew. By 2006, a large multigroup study of the mutant genes in breast and colorectal tumors had found that 189 different genes are frequently mutated in these tumors, and that any given tumor cell has an average of 90 mutated genes. Other studies have determined that the two so-called BRCA genes, flawed versions of which were linked to breast cancer in 1995, turn out to play a role in only 5% of breast-cancer cases.
The simple fact is we still just don’t know very much about genes, says Craig Venter, who famously spearheaded the push to sequence the human genome, founded Celera, and remains a driving force in genetics research. “We don’t know what most genes do, and we certainly don’t know what the variations are in most people. The idea that we can design custom drugs around genes, or change genes, is just silliness and science fiction.”
“Gattaca got it totally wrong,” says Bryan Walser, CEO of gene-discovery company Perlegen Sciences, in Mountain View, California. “In the movie, genes have 100% penetration,” meaning that if you have a flawed gene, it’s certain you’ll get the disease it’s associated with. For most major common diseases, he explains, specific genes are almost never associated with more than a 20% to 30% increased chance of getting sick. Indeed, the notion that a small number of genes represents a large component of the risk for a particular disorder has simply turned out to be untrue for almost all major illnesses. And the weakness of these correlations extends to other attributes as well. The gene most strongly linked to intelligence accounts for less than 0.4% of the observed variation, while the top six intelligence genes together predict 1% of the variation. A 2009 study of about 6,000 people came up with a technique for predicting a person’s height by looking at the 54 height-related genes; the results turned out to be one-tenth as accurate as averaging the heights of both parents and adjusting for sex, a technique introduced in 1886 by statistician Sir Francis Galton.
Even the very presence of a given gene is a rabbit hole of confusion. Genes can be “turned off” so that they might as well not be there, or partly turned on so that they contribute only weakly to the disease risk. Asked what sort of “switches” turn genes on and off, Ahituv fired off a partial list: one another, the environment, various poorly understood forms of the DNA-like molecules called RNA, and even strips of junk DNA. In fact, Ahituv studies only junk DNA. “After I got my PhD, around the time the genome was being finished, I had a feeling the junk would turn out to be important, so that’s what I’ve focused on,” he says. Unfortunately for our odds of living to 120, that hunch turned out to be right on the money. “It’s very discouraging, but we don’t have any kind of code for understanding junk DNA,” Ahituv sighs. “I can find the switches, but I don’t know what they do. There are switches for the switches, and switches for those switches. It’s endless.”
Junk DNA isn’t just a minor hitch in the gene-disease picture — it accounts for as much as 80% of the genome’s influence over disease. There are other complicating factors: Genes can appear in single form, or in multiple copies that increase their influence in ways most gene tests don’t detect. They can pop out of their slots in the genome, replicate themselves, and then reinsert themselves somewhere else in the genome. They can flip around backward, which also isn’t detected by gene tests, even though it can alter the resulting proteins. Throw in the fact that viruses can insert their DNA into the genome, and that any protein made by a particular gene can do different things in different people, or different things in the same person at different times depending on what else is going on in the body, and you’ve got complexity of such staggering breadth and depth that scientists like Ahituv have come to recognize that they’ll be lucky to make a small dent in the task of sorting it all out in their lifetimes. “It’s job security for researchers,” he says.
Just about everyone involved in genetic research and biotech seems to agree we were all sold a bill of goods. “The genome was sold to the public as a transcendental leap into the banishing of disease,” says Lev Osherovich, a biology researcher currently on sabbatical as a senior writer for biotech industry Web site SciBX. Not everyone agrees who was guilty of the hardest selling. NeuroInsights’s Eeckman says it was biotech entrepreneurs looking for investors and pharma deals. Some biotech executives say it was researchers eager to loose enormous rivers of government and industry funding. (“They overpromised so they could get the genome done,” says Perlegen’s Walser.) Researchers blame science journalists.
For scientists, the fact that the genome has raised far more questions than it has answered is a treasure trove of opportunity. The National Institutes of Health has funded roughly 280 new studies that each comb through multiple genomes; one new project alone has set out to map the genomes of 1,000 people. About 35 million strips of DNA have already been catalogued in one form or another as being common to all people, and the number is growing steadily. But “we end up with a lot of false discoveries, and we’re missing a lot of important ones,” says Walser. Solidly verifying most potentially useful gene links would necessitate poring over at least 10,000 genomic samples, and rarer links would require samples from more people than live on the planet.
The one corner of the genome-focused biotech industry that’s thriving is the one churning out equipment and services to support researchers in their endless hunt for gene links. I visited gene-analysis equipment vendor Affymetrix, where I got to see the state of the art, which strangely resembles a small collection of kitchen appliances, including toasters, rotisserie ovens, and dorm refrigerators, albeit versions that cost tens of thousands of dollars. Affymetrix’s latest offering, the Gene-Titan, can simultaneously pore over 96 DNA chips (like electronic chips, but with tiny chambers for DNA samples instead of circuitry), analyzing nearly 800 a week. It helps explain how the cost of analyzing a full genome has dropped from that $3 billion in 2003 to about $10,000 today — beating Moore’s Law by a factor of about five. Meanwhile, the fact that disease turns out to be controlled not just by genes, but also by various types of RNA, junk DNA, and more has led to an explosion of entirely new projects aimed at cataloguing as many of these “epigenetic” links as possible. The New York Times has already run a special section on the miracles that the new science of epigenetics will be bringing us. “The hype about the genome did everyone a huge disservice,” says Affymetrix chief medical officer Richard Hockett. “Now the next generation of studies is getting the same treatment.”
Many say it’s not fair to claim that the genome hasn’t paid off, insisting that all sorts of marvelous new drugs are still in the decade-long drug-development and testing pipeline. It’s true that the Human Genome Sciences lupus drug suddenly looks promising, and Robert Weinberg, a pioneering cancer researcher at MIT, told me he expects encouraging results from a melanoma drug targeting a protein identified through genomic research. But other specific examples are hard to come by. While a few gene-related drugs have made it to market, including Gleevec, Iressa, and Tarceva, all were developed in the pregenome days with relatively crude gene-finding tools and have proven effective only for a small fraction of patients — and even in these patients, the disease often becomes resistant to the treatment. Among the dozens of genome-inspired drugs in testing, the chances that more than a couple will pan out are slim.
As Celera CEO Kathy Ordoñez says, “The fundamental steps in drug development don’t change because of genome studies.” Those steps screen out the vast majority of candidates: Most fail in animal testing, and of those that don’t, and that a drug company backs all the way through human testing to the tune of hundreds of millions of dollars, only about one in 10 will get through all the hoops. Genome-based treatments are in the drug pipeline in the same way that the Cleveland Indians are in the World Series pipeline.
Okay, say others, forget new drugs — what we’re getting out of genomics is “personalized medicine.” That is, by having our personal genes read, we’ll find out what diseases we’re at risk for and which drugs will work best for us. There are some 1,500 gene tests available today, about 1,000 of which are sold directly to consumers for up to $1,000. And there have been a few modest successes in personalized medicine: Celera offers a gene test that can help determine if a patient ought to be on cholesterol-lowering “statins,” Perlegen sells one that helps identify good candidates for drugs that can lower breast-cancer risks, and other tests can aid in predicting who will be most likely to benefit from certain experimental Alzheimer’s drugs or who would be harmed by a blood-thinning medication often given to heart-attack victims. But these are exceptions, intended for use by physicians. Given how hard it is to reckon the link between individual genes and disease, how much insight should we expect from the tech guy at a mail-order Genes ‘R’ Us? The Federal Trade Commission answers that question on its Web site: “Some companies claim that at-home genetic tests can measure the risk of developing a particular disease, like heart disease, diabetes, cancer, or Alzheimer’s. But the FDA and CDC say they aren’t aware of any valid studies that prove these tests give accurate results.” Even the much ballyhooed 23andMe, backed by Google, received words of support from only 2 of the more than 25 scientists and biotech executives I spoke with — and that support was tepid: Stanford’s Altman, who is a paid adviser to the company, says a 23andMe readout could provide “somewhat interesting information,” and Venter calls it “a good introduction to the concepts, if you understand how limited it is.”
“The question is, What does knowing about a gene give you that you don’t already know?” says Jay Kaufman, who heads product marketing at Affymetrix. Why get tested for genetic risk of heart disease if your doctor has already put you on statins because you’re overweight and have high LDL cholesterol? What can your doctor do with the fact that you carry a gene that puts you at higher risk of Alzheimer’s? Who needs a gene test to find out they’re at risk for obesity, or alcoholism, when they can just look in the mirror or count their empties?
None of this is to say we shouldn’t have bothered with the genome, or that we should stop working on it now. But we shouldn’t base our decisions to invest in the science or in the biotech that comes out of it on an incomplete understanding of how long a task we’re facing.
And we needn’t be quick to see the failure of the genome to spill our medical secrets as a terrible thing. When physicists figured out quantum mechanics in the early 20th century and learned from it that everything that happens is a matter of chance, Einstein was famously among those who expressed horror at the notion that nothing was fully predictable. “God does not play dice with the universe,” he grumbled. But others realized that quantum mechanics frees us from the restrictive notion that our lives play out in the only way they possibly could. Maybe the fact that our medical fates are still mysteries to science is something we should be thankful for.
Besides, there really has been a payoff from the genome, if an indirect one: In the vast majority of cases, individual genes apparently don’t influence your destiny as much as, or at least any more than, your behavior does. So lose weight. Get some exercise. Trade in the cheeseburgers. Breathe clean air. And for God’s sake, don’t smoke. It’s pretty much the same advice your great-grandfather got from his doctor. I bet it’s the same advice your great-grandchildren will get from theirs.
But I guess we’ll keep hoping for that magic bullet. When I was in Nadav Ahituv’s office, I noticed his window looked out on an enormous stand of trees, a surprising sight in a building that fronts a busy San Francisco street. I asked about it, and he told me they were eucalyptus trees. But then he added he wouldn’t be seeing them much longer and pointed out a fleet of construction vehicles lined up near the trees. “They’re putting up a stem-cell lab,” he said. “Every major university in California is getting one.” I made a note to check back 10 years from now to see how that’s working out.
David H. Freedman’s next book, Wrong, about why expert opinion is often flawed, will be published by Little, Brown in 2010.