In an ideal marketplace, customers who are dissatisfied with their products can easily move on to the next one. But for many microarray core facilities around the world it often seems that once you make the initial investment into a chip platform, you are anchored to that platform for life.
"Researchers are generally stuck with platforms that they adopted early," said Mark Geraci, director of the Gene Expression Facility at the University of Colorado Health Sciences Center, one of the largest core labs in the US that runs experiments on the Affymetrix platform.
Geraci, for example, has found it easy to stick with Affy: The company has courted his lab, kept it stocked with up-to-date technology, and offers what he considers a pricing structure that is "comparable to other platforms." Plus, Affymetrix is the platform most of his colleagues use, he said.
"Scientifically, when you put Affy in PubMed you get close to 3,000 scientific articles using the platform," Geraci explained. "If you put CodeLink in, [for example], you get about 14."
Carl Borrebaeck, director of SweGene's microarray core lab in Lund, Sweden, said that his facility had evaluated other platforms but that it had decided to stick with Affy.
"We have evaluated some other platforms, such as [CodeLink], et cetera, and none of them were competitive enough to convince us to change over to a new technology. They worked fine, but so does the Affymetrix platform, which is very robust," Borrebaeck said.
In addition to these factors, if there is one thing that is preventing labs like Borrebaeck's and Geraci's from switching microarray platforms to Agilent, GE Healthcare, or Applied Biosystems, it is the significant investment they have made in generating data.
"It would be enormously hard [to switch]," Geraci conceded. "Because all of our data, which is now somewhere around three terabytes of data, is 95 percent on the Affy platform."
But what of the other five percent? Geraci said that 18 months ago his facility started running assays on GE's CodeLink platform, and, to a lesser extent, on Agilent's arrays.
"The reason [that we started] with CodeLink was the quality of the low signal data, which is frequently absent on the Affy platform," said Geraci.
Besides, Geraci said, he was already equipped with the necessary tools to work with CodeLink - such as GenePix software and BeadScanner reader - so the cost of adding another platform was negligible.
"If you move from an open-ended platform to Affymetrix it costs you a lot, because setting up the system is around $250,000 to $300,000," said Naftali Kaminski, who heads the Simmons Center for Interstitial Lung Disease, a core lab in Pittsburgh.
"If you go the other way around it's actually not very expensive because a good scanner is something between $50,000 to $100,000," he said.
Kaminski knows the price of switching platforms because two years ago he decided to forego the Affymetrix platform he had been using since 1997 to use CodeLink bioarrays.
He cited several reasons for making the decision, but in the end he was more moved by the technology than price incentives.
"I think that all of them come to the same price in the end," Kaminski said. "I think it was mostly the fact that I thought that Affy was more dependent on normalization, you could never really see the signal, and the fact that I wanted an open platform," explained Kaminski.
Kaminski also said that a third reason for switching was that there was another core lab in Pittsburgh running Affy arrays, and he therefore felt no need to stick with Affy, even though it is the dominant platform in the market.
"If you have a big core facility or institute [in a region], you sort of say, 'Why waste $250,000 in setting up a second one?' So I wanted to have a complementary system," he said.
Unlike Geraci, Kaminski said he had no trouble discarding his old Affy data because at the time he decided to switch he couldn't compare it to data generated from new Affy chips, let alone to CodeLink or Agilent arrays.
"That's actually one of the problems with Affy," said Kaminski. "Whenever they came out with a new batch of chips it was hardly at all comparable. So it wasn't like we were losing a lot."
"It's almost impossible" comparing older Affy data against newer Affy data, said Geraci.
"The probe sets are different, so when they changed the location of the probe set within the genes, they changed the affinities for hybridization and the efficiencies of reverse transcription," he said.
Affymetrix did not reply for comment on this article.
Despite this challenge, however, both Kaminski and Geraci agree that the data produced with Affy has become more stable in recent years. Meantime, Kyle Serikawa, director of the Gene Expression Facility at the University of Washington in Seattle, who started using Affy in 2001, said he hasn't had problems comparing older Affy data with newer Affy data.
"Affy allows great comparability among arrays - even arrays done several months apart," said Serikawa. "They also have very good service and the widest selection of products."
Still, in 2003, Serikawa said his facility began to offer customers Agilent arrays, as well as Affy chips, and today the facility's arrays are about 80 percent from Affy and 20 percent from Agilent.
One of the downsides to Affy, said Serikawa, was the large cost of obtaining the equipment, while for Agilent there was less of an "upfront investment in equipment."
Shawn Levy, who runs the Microarray Shared Resource at Vanderbilt University, said he thinks the "dilemma" of cross-platform analysis in the microarray space is exaggerated.
"All platforms are comparable if the data is analyzed correctly," he said. "Some of the recent press that indicated that cross platform comparisons show no comparability between data sets run on different platforms is over-hyped and incomplete."
Levy knows a bit about cross-platform analysis because he has managed to offer users of his facility gene-expression platforms made by Affymetrix, CodeLink, Agilent, and, this year, ABI.
According to Levy, a properly performed and analyzed experiment will yield highly correlative results. Yet while he has his methods down pat, others, like University of Colorado's Geraci, are using different methods to compare data
In Geraci's case, his lab compares older Affy data with newer data by running both through Stratagene software and then using the Stratagene results as a "common denominator" for analyzing data.
"It's the most effective way to cross platforms," he said. "So rather than going to PCR reactions, we just run arrays."
The idea of a metric or a standardization of data isn't new in the industry. Most experts agree it is a good idea, though little has been done to give lab directors like Kaminski, Geraci, Levy, and Serikawa a universal method for comparing data that will make them less hesitant about moving from platform to platform.
Some microarray users, like Francis Barany, a professor of microbiology at Cornell University, think the current problems associated with switching platforms are just the growing pains that accompany any developing industry.
"This is part of its evolution, this is part of its growing process," he said. "In the early days there are no standards, but as the field becomes more rigorous there becomes a demand for standards."
Barany, who is also a consultant for ABI, said that younger companies in the space would probably be the ones that would lead the way for standardizing data and creating the kinds of metrics that would enable microarray users to switch from platform to platform.
"It would appear to me that those with the greatest incentive to do so would be companies that currently have platforms that are not dominant and would like to get a larger market share. Those companies would have an incentive for developing conversion tables," he said.
Still, while Barany said there is a "crying need" for such standardization, neither he nor any of the other experts questioned by BioArray News said they were aware of steps being taken by any array manufacturers to standardize data at this time.