The Gene Machine

By Dick Thompson

Craig Venter has no shortage of rivals who would love to see him fail--especially among scientists at the Human Genome Project, the multibillion-dollar government-sponsored effort to map every one of our 100,000 genes. When the millionaire molecular geneticist announced in 1998 that his company, Celera Genomics, would do the job in a third of the time at no cost to the taxpayer (thereby making the Genome Project seem like a wasted effort), the scientific community was split into two camps--one group of researchers hoping he could make good on his promise, the other predicting he would fall flat on his face.

So it was with sweet delight that Venter announced last week that Celera had completed a rough draft of the genome months ahead of schedule and that it was on track to wrap the project up as early as this summer. Celera scientists now have 90% of the genome in their database. They have also captured 97% of all the known human genes and discovered tens of thousands of new ones--including hundreds for previously unknown neurotransmitter receptors and at least one new kind of interferon. "This is not only a monumental moment in Celera's history," Venter proclaimed in a webcast news conference last week, "but also in the history of medicine."

Hyperbole? Sure. But it may also be true. It is conventional wisdom among scientists that having the complete genetic blueprint for humans will transform medicine--with new classes of drugs, new kinds of treatments and new definitions of disease based not on overt symptoms but on underlying genetic defects. Armed with a patient's genetic data, doctors may someday be able to diagnose diseases before they occur or prescribe medications custom-made to fit each patient's unique genetic profile.

How can we reach this promised land? Here the gene mappers part company, the scientists with the Genome Project having gone one way and Venter quite another.

The Genome Project, like the proverbial tortoise, took the slow and steady route. Scientists first divided the full complement of human DNA into 22,000 segments, each 150,000 letters long. The positions of these segments were carefully mapped, and then each was cloned several times. Those cloned segments are now being decoded by automated gene sequencers, and the process repeated several times to ensure accuracy and close any gaps in the coverage. Because each segment was mapped before cloning, the decoded segments can be easily fitted back into their original position in the completed genetic map.

Venter is using a faster but more risky method he calls "whole genome shotgunning." He clones a genome several times and then blasts the clones into 60 million bits, each between 2,000 and 10,000 letters long. Each fragment is then fed into a high-speed decoding robot. The next step, for Venter, is the most difficult. His robots e-mail their results to Celera's giant central database (said to represent more concentrated computing power than anywhere outside the Pentagon). These computers are using a sophisticated program to reassemble the genome fragments into the familiar 23 human chromosomes. The whole process can be compared to making confetti out of a stack of encyclopedias and then painstakingly reconstructing each page.

Venter's rivals have raised questions about how complete and accurate the finished maps will be. Venter has acknowledged that Celera's will contain some gaps, especially in the so-called repeats--long stretches of DNA with virtually identical sequences. Genome Project scientists once argued that because they have an easier time sequencing the repeats, their finished map would be more complete and error-free. They have since discovered that many regions throughout the chromosomes are unreadable even with their technology, and they are now forced to acknowledge that their product will have large gaps as well.

The best reason to keep the public Genome Project going, its scientists say, is that it is, well, public. Alarm bells rang last year when Celera announced that it had filed provisional patents on hundreds of newly discovered genes--a list that by last week had grown to include thousands of genes. Venter has pledged that he will eventually give away the completely decoded genome and make his money by selling the computer services needed to make sense of it. For now, however, he is charging $5,000 to $5 million a year to wade through his data.

The Human Genome Project, by contrast, publishes its results on the Web, free to all comers. One of its biggest users, it turns out, is Celera. Of the 90% represented in Venter's rough draft, only 81% was sequenced by Celera's robots. The rest was taken from the public website.