Chips Ahoy
Zina Moukheiber, Forbes Magazine, 11.13.00
forbes.com
IN A BUILDING STILL UNDER CONSTRUCTION, technicians wearing hair nets hunch over machines the
size of a small refrigerator. A robotic arm spits bits of DNA onto glass sheets at the rate of 8,000 bits per second.
A dicer cuts the sheets into 3-by-1-inch wafers.
This is not your bootstrapped biotech company. The scientists making the DNA chips labor for Agilent
Technologies, the $8 billion (sales) spinoff of Hewlett-Packard. In April it moved into a former HP building in
Santa Clara, Calif. and is converting it into a 15,000-square-foot DNA chip factory.
"There's a revolution in genomics, and we have a rich technology portfolio that we can apply to this," says Ned
Barnholt, Agilent's chief executive.
With the mapping of the human genome, DNA chips are shaping up to be a dominant platform for analyzing gene
function. Scientists use them to compare genepatterns in healthy and diseased cells to learn both how diseases
work and which genes might pose targets for new drugs. Down the road, DNA chips could become a tool for
detecting disease in patients and letting doctors see test results in only hours instead of weeks. Diagnostics will
detect whether, say, a woman has a defect in her p53 gene, predisposing her to breast cancer, (see "Divining
Disease" ). Sales of DNA chips could hit $300 million this year and grow fivefold by 2006. The market is ruled by
Affymetrix, the self-styled Intel of DNA chips with a 60% share. But now Agilent is knocking, as are other
electronics makers with deep pockets and solid reputations. Motorola, Corning and Hitachi have formed "life
sciences" divisions and are about to produce their own DNA chips.
Affymetrix pioneered DNA chips in the early 1990s. This year its chip sales will double to 200,000, and it will run
a loss on estimated sales of $190 million; no matter, its market value is $3 billion.
The electronics rivals all plan to produce chips that are cheaper and more flexible than the Affymetrix line. Making
these biotech tools isn't that far removed from their expertise in semiconductors, ink-jet printers or optically pure
glass.
Agilent, based in Palo Alto, Calif., already had a foothold in the pharmaceuticals business, selling instruments to
separate and measure proteins. "The next logical step was to go upstream in the information path," says Barry
Willis, who spearheaded Agilent's DNA chip project. The company's burgeoning Life Sciences division has tripled
its staff to 170 this year, as Agilent looks to maintain its recent growth rate of 28% a year.
For Agilent, the new foray will mean taking on an old friend. In 1994 Agilent's Willis pitched the idea of jointly
developing DNA chips to Affymetrix founder Stephen Fodor, offering ways to help cut Affymetrix's production
costs. The two companies had signed a deal involving Agilent scanners, and talked about Agilent selling Affymetrix
chips. But Fodor declined Willis' offer (he won't disclose details). The two companies scrapped the chip resale
proposal. Affymetrix is still under contract to buy Agilent scanners until 2003, but suddenly Fodor's supplier could
become his chief competitor.
Affymetrix makes its chips using photolithography, the same way semiconductor chips are made. Agilent is pursuing
a newer technique that uses ink-jet nozzles to squirt pieces of DNA onto a glass wafer.
Affymetrix's process starts with a glass wafer the size of a dime and builds up to 25 layers of the four-letter
alphabet of DNA, one letter at a time, shining a light through a mask to bind each layer at exactly the same place.
The short DNA sequences go into microscopic channels, or "spots." The chips are packed with information;
400,000 spots on a chip express 12,000 genes.Because they are complicated to make--a different mask is needed
for every letter of DNA--Affymetrix's chips are expensive. A chip with 12,000 human genes goes for $2,000, plus
a hefty upfront fee. They are not reusable, so scientists run through a slew of them. One experiment can cost more
than $150,000.
"Affymetrix chips are out of consideration, because of their cost," says John Quackenbush, a molecular biologist at
the Institute for Genomic Research. He uses homegrown chips for his work on genes present in colon cancer.
Agilent's cheaper nozzle technique owes to the partner it picked up after its alliance with Affymetrix unraveled:
Rosetta Inpharmatics in Kirkland, Wash. It owned the rights to the ink-jet method, and Agilent added
precision-printing know-how from its HP days. Much like printing a picture pixel by pixel, the process requires
positioning ink-jets very precisely over the same tiny spot, again and again, to build DNA sequences. (see below).
Agilent can fit only one-sixteenth the number of spots on a chip as Affymetrix, but ink-jet printing is more flexible
than photolithography. Agilent can program its machines to print however many spots in whatever sequence the
client needs and deliver the chip, all in two weeks.
"Like print-on-demand, this is chip-on-demand," says Vincent Dauciunas, who heads business development. The
one-size-fits-all Affymetrix chip takes at least two months to redesign, because new masks have to be built. Agilent
chips are cheaper, too. An 8,500-gene chip goes for $400. By the end of the year it will sell a 25,000-gene chip
for around $1,200.
Paradigm Genetics in Triangle Research Park, N.C. recently picked Agilent over Affymetrix. Paradigm looks at
gene characteristics in fungi and rice for clients such as Monsanto and Bayer. It expects to use 500 chips per month
and to save $1 million. Says project manager Phanos Maroglou: "It's ten times less expensive to use Agilent's chip
design."
Agilent still must prove its chips are reliable. Corning, meanwhile, began selling its own yeast-gene chips in
September and plans to offer human DNA chips holding 10,000 genes later this year. Customers will benefit: Two
years ago a 6,000-gene chip from Affymetrix cost $2,000; that price now buys 12,000 genes. More to come. |
