By John S. MacNeil
Nestled in the leafy confines of Rockefeller University’s campus on the Upper East Side of Manhattan overlooking the East River, Greg Khitrov’s microarray laboratory is a well-heeled but otherwise typical academic core facility. There’s space for printing custom arrays, an Affymetrix platform, a room with three Pentium workstations for data analysis — even a few laboratory technicians to look after. And like most core facility directors, when not overseeing day-to-day operations, Khitrov parries vendors who come offering new gadgets that could help him serve his customers.
Now Khitrov has a complicated new purchasing decision to make. Applied Biosystems, Agilent Technologies, Affymetrix, and other vendors are beginning to offer the whole human genome — as it’s now understood — on one microchip, and Khitrov is smack in the middle of the marketing maelstrom. Already, ABI has called to gauge his interest in its early-stage system (“not interested” in switching platforms, he says), and he’s spoken with Agilent about its new whole-genome chip that hits the shelves this month.
Khitrov’s conundrum is one many microarray users will face over the next few months. Because many employ the technology to generate new hypotheses, they’d prefer to scan the whole genome in one fell swoop. The challenge for people in Khitrov’s position is to pick the tools that accomplish the task most efficiently. A microarray hosting the whole human genome would make experiments faster and more reproducible, conserve sample and reagents, and, in theory, produce higher quality results, researchers say. But how readily users adopt the new technology depends fundamentally on one variable: cost.
“It’s definitely a big deal — it’s definitely important to have all these things on one array,” says Stan Nelson, director of UCLA’s microarray core facility. “The issue is to do that without having to pay a huge price.”
In fact, the most significant news to microarray laboratories may not be the launch of the whole-genome chip product itself, but the number of companies that are attempting to offer it. Commercial microarray chips are expensive. But competition will chisel away at cost, making even the most powerful arrays more accessible. “Every time platform coverage goes up, the cost per gene goes down,” says Michael Geraci, director of the microarray core facility at the University of Colorado Health Sciences Center in Denver. “For a lot of reasons, people are excited.”
For years, biologists have dreamed of the day when it would be a simple matter to carry probes for the entire human genome on a single 1” x 3” slide. That microarray producers can now accomplish this feat, just 12 years after Affymetrix co-founder Stephen Fodor and his colleagues first described the technology in Science, is a testament to the ingenuity of researchers — in academia and industry — responsible for upgrading chips and platforms.
But it’s not as if microarray users didn’t see this coming. Affymetrix and Agilent each already sell two chips that represent their versions of the human genome, and consolidating that content onto one chip is not a huge technical hurdle, says Andrew Brooks, director of the functional genomics core facility at the University of Rochester. Furthermore, because the whole-genome chips now entering the market rely on the current annotation of the human code, initial versions will be significantly less than comprehensive. In fact, researchers aren’t even sure how many human genes actually exist.
The more profound implications for biology lie in what chips of increasing density will deliver in the future. In the coming months and years microarray researchers and vendors hope to pack truly representative probe libraries on single chips, such as for every human gene plus all splice variants. Also not far off is the day when vendors can offer one array containing probes for multiple microbial genomes, or a single chip that comprehensively covers larger genomes such as rice and wheat.
Without a doubt, the whole-human-genome chips soon to become available by catalog are intriguing. Single chips containing probes for the entire human genome are already available in customized form from the likes of Nimblegen Systems, which offers a microarray analysis service, and Affymetrix (which announced last month that its custom chips can accommodate up to 61,000 transcripts on a single array), and Agilent may be the first to market a mass-produced chip containing oligonucleotide probes for 36,000 human genes and transcripts, with an additional 8,000 “housekeeping” probes. And, to the presumable dismay of its competitors, even ABI has entered the field, announcing in late July that it too plans to offer a whole-human-genome chip and associated analysis platform.
How researchers respond to each of these companies’ offerings will depend largely on the pricing and positioning strategies vendors reveal in months to come. Here Genome Technology delves into the issues you have to sort through to decide which, if any, of the new whole-genome chip platforms to install in your lab.
What’s a whole-genome chip, anyway?
Perhaps the most immediate technical question any savvy chip user will ask is, “What exactly do vendors mean by ‘whole genome’?” Several researchers who spoke to Genome Technology questioned how early versions can be truly representative, since the genomics community is still undecided as to the exact number of human genes. “There are only about 21,000 genes annotated in Ensembl, and there’s evidence there’s a good deal more genes than that,” says John Quackenbush, an investigator in microarray informatics at TIGR. In fact, this past summer the powers that be in the genomics community settled on 24,500 as the currently confirmed number of protein-coding genes. But the vendors closest to offering a catalog whole-genome chip all say they can build probes for significantly more than that number. What are their sources?
Agilent says it will design oligo-nucleotide probes based on the 36,000 genes and transcripts — the definition of gene varies according to the researcher, Agilent says — it has identified with the help of Incyte’s LifeSeq Foundation database, and with information from public databases such as Ensembl, RefSeq, and UCSC’s GoldenPath. To synthesize probes for the 39,000 transcripts from 33,000 genes represented on the U133A and U133B human genome sets — an updated version of the content will appear on the first version of its whole-genome chip — Affymetrix says its team of bioinformaticians relied only on public data.
And according to its July press release, Applied Biosystems is relying on public databases as well as the Celera Discovery System to assemble the “more than” 30,000 genes it will target with its whole-human-genome array. (ABI declined to be interviewed for this article.) Nimblegen, which only offers its whole-genome chip as part of a microarray analysis service, says it relies on GenBank and GoldenPath to derive its 38,109 genes.
Stay loyal or take a leap?
Rochester’s Andrew Brooks says he’s more interested in a vendor’s ability to explain the significance of the gene expression data collected with its platform than he is in packing a lot of probes onto one chip. “Having the entire human genome could potentially be a disadvantage, unless the informatics exist to help us understand that information better,” Brooks says.
To that end, Agilent says it provides customers free access to the gene annotation data in Incyte’s LifeSeq Foundation database; Affymetrix provides free access to its NetAffx online data analysis center; and ABI says it will provide an “industry standard” database, complete with gene acronyms and names, cross references for gene identification, gene ontologies, and protein characterization data. For an additional fee, ABI also offers access to the Celera Discovery System.
A platform comparison based strictly on the strength of its informatics infrastructure may show ABI having the advantage, says Brooks, who signed a non-disclosure agreement with the company. While ABI’s experience with microarray platforms is limited, it can rely on Celera’s genomics expertise for annotation, and it can provide related tools such as Taqman reagents, which could be useful for validating micro-array-derived gene expression results, notes UCLA’s Nelson.
Says Brooks: “I don’t think ABI has to do it that much better … [or] much differently. I think that, like any other high-throughput technology, they’re leveraging their expertise on understanding the human and the mouse genomes through their sister company Celera, giving us a better understanding of how to interpret this information. To me that means more.”
But to those who have already invested in one closed platform such as Affy’s, dishing out for ABI’s system — which requires an ABI scanner — might be out of the question. ABI has yet to reveal its pricing plans, but Affy has an installed base of 860 platforms worldwide. Although Affy won’t discuss pricing, some users have paid at least $200,000 for its platform, an investment that may induce customers to remain loyal purely out of financial necessity. “Academics don’t have that kind of money,” says Chris Barker, director of the Gladstone Genomics Core at the University of California, San Francisco. “[ABI is] saying, ‘Throw away your investment.’ … One or two years ago they would have had a much easier way of getting into the market.”
Affy users — and others who’ve relied predominantly on one platform — must also take into account the investment in the data they’ve collected. Studies involving multiple cancer patients require researchers to compare new data with old data, a task that becomes entangled in compatibility problems if scientists decide to switch platforms during an ongoing investigation, says Gavin Sherlock, a Stanford microarray informatics researcher. Furthermore, a database configured to store data from one platform might have to be reconfigured to accept data from a new one.
Brooks says, “The advantage that Affymetrix and companies that have been doing it longer and doing it well [have] … is that there’s a huge amount of historical data that exists. It’s going to be very difficult for any one lab, or any one program, to completely switch gears without sacrificing all of the time, money, and energy they’ve invested in that data.”
Open or closed platform?
And then there’s the question of open versus closed platform. Microarray users considering purchasing Agilent’s whole-genome chips may still have to worry about data compatibility, but they won’t have to bother buying a scanning system specific to its microarrays, as is the case for Affy and ABI. For many researchers — especially in academia — who spot and analyze “homebrew” chips, Agilent’s open platform arrays are attractive because they can use whichever commercial 1” x 3” slide scanners they currently have on hand.
An open-platform system has other advantages as well, particularly if you’re looking to maximize your ability to sample new products at minimal cost, says John Feder, a group leader in functional genomics technologies at Bristol-Myers Squibb who has also invested in an Affy platform. “An open system is of paramount importance to us because they’re amenable to integrating new technologies,” he says. “We try to remain flexible — we don’t want to be cornered into doing one thing.”
On the other hand, working with a scanner optimized for a specific chip, as are Affy and others’ closed platforms, should help eliminate some sources of experimental error. In addition, Affy says its platform also provides one-stop shopping and ensures product compatability and support.
OK, what’s it gonna cost me?
Unfortunately for researchers looking to price shop, vendors are only hinting at what they plan to charge. Agilent and Affy, which say their chips will be available before the end of the year, have both indicated that their whole-genome microarrays will cost less than what two cost today, but more than one. ABI won’t comment on the cost of its system, which includes a chemiluminescence scanner.
The question for you if you’re cost conscious (and who isn’t?) is whether the new chips will be affordable, particularly in comparison to the cost of producing spotted arrays in-house. Already, Affy customers complain about the vendor’s prices, in some cases even declining to run their samples on the U133B chip — the less well-annotated of the human genome set — to save money, says UCSF’s Barker. Several academics who spot their own arrays at a cost of $200 per chip say prices for Affy’s whole-genome arrays would have to come in at $400 for them to buy them en masse. “At $400, that would be a good start,” says George Watts, director of the microarray core facility at the Arizona Cancer Center in Tucson. “At $150 to $200, you’d really have something there.”
Given the advantages in quality and reproducibility inherent in a mass-produced array, these customers might seem to be asking too much. But increasing competition should, in the long run, lead to better deals for microarray users. Some researchers even predict that spotting one’s own arrays will soon go the way of the slide rule: “It’s simple to think how to do it, but hard to do it well,” Ernest Kawasaki, head of the microarray facility at the Center for Cancer Research at the National Cancer Institute, says of producing one’s own arrays. “Now you’re going to have three or four companies [manufacturing arrays], and I’ll be out of a job. I won’t be sorry to see it go; it’s not actually that much fun.”
Who else is in the game?
With competitors stumbling over each other to position themselves as the whole-genome chip vendor of choice, you might not have to wait long for prices to start dropping. ABI’s announcement at the end of July heralding its entry into the microarray market seems to have caught Agilent and Affy by surprise (ABI scooped Agilent on its launch, which was planned for early October), and the race to bring a chip to market may have encouraged Affy to speed its introduction. “To me, all that’s happened is that ABI has forced Affy to accelerate their move [to a whole-genome chip],” says Barker. “I don’t think Affy planned to do it so quickly.”
Meantime, one company, Nimblegen Systems, based in Madison, Wis., already boasts its own whole-human-genome microarray chip, although it’s not for sale. The company instead offers a microarray analysis service, and customizes each array using micromirrors and photo-deprotection chemistry to build 60-mer oligos representing up to 38,100 different human genes, with an average coverage of five probes per gene, says Emile Nuwaysir, Nimblegen’s director of business development. “Our new customers are choosing this over other designs,” he says. “There are no other human arrays available.”
Others are waiting in the wings. Illumina, known for its bead-based arrays, says it also has plans to construct a whole-human-genome offering. The company initially focused on the SNP genotyping market, but its technology, which relies on conventional fluorescence scanners, can accommodate 6.7 million three-micron features on a single chip, making gene expression the next logical application for its technology, says CEO Jay Flatley. Even accounting for a 30-fold bead redundancy per probe, the company’s technology still has more than enough density to represent 30,000 genes on a 1”x 3” glass slide, he adds. “I definitely think we’ll be in the game,” says Flatley, although he won’t say how soon. (Besides, he says, “everybody else that’s being talked about, they’re all “A” companies — ABI, Agilent, Affymetrix — there needs to be another vowel in the mix!”)
One “A” player noticeably absent from the chorus of competing claims is Amersham Biosciences, which purchased the CodeLink microarray platform from Motorola in August 2002 for $20 million. Although the CodeLink business has yet to make money — Amersham’s mid-year 2003 financial report put its net expenditures for CodeLink at $7 million — Amersham has not given up on challenging Affy’s market dominance, according to a report in GT’s sister publication Bio-Array News. The company declined requests for interviews for this article.
Meaningless milestone or must-have technology?
As vendors continue to improve chip density, it’s likely that researchers will look back on these first versions of whole-genome chips the way computer users reminisce about the Apple IIE. Affy Chief Commercial Officer Trevor Nicholls says his company plans to produce a set of chips containing all the exons in the genome within the next year. And that’s just a start, he adds. Affy has demonstrated that it can build chips with features of less than 10 microns, and in a research setting has synthesized features under two microns in size. This kind of density should allow Affy to place much larger genomes, such as rice, or perhaps several smaller genomes, on one chip.
In essence, a single chip that merely consolidates the content of multiple chips available today “is just a milestone, because nobody is talking about putting the ‘genuine’ whole genome on a chip,” Nicholls says. “It’s just a representation of 30,000 to 50,000 genes, and really you want to move beyond that to include all the exons, splice variants — even junk DNA.” To that end, Nicholls says Affy has plans to start “looking at a broader view” of the transcriptome, splice variants, and other gene regulating factors within “a two-year time scale.”
Other vendors, including Agilent, make similar claims to the ultimate power of their chips, but there is some debate about how densely one can print 60-mer oligos on an array. Affy claims that because its arraying technology is based on photolithography, its chips have the potential to target a far greater number of transcripts than what a robotic ink-jet printing system could ever accomplish. Agilent, for its part, says there is “no significant limit” to the feature density chips, but that “our decisions about increasing density are guided by quality of experiment results.”
Regardless of the imperfections that early versions are bound to contain, most everyday microarray researchers are itching to get their hands on a whole-genome chip. While the advances in density and informatics that led to its introduction were primarily incremental, the advent of a whole-genome microarray represents a significant milestone in the evolution of microarray technology, as well as the industry as a whole.
To be sure, whichever company is first to launch a version of a whole-genome microarray won’t necessarily be the one that ultimately wins the most customers. “It’s not like a drug, where being first to market determines your profit,” says UCSF’s Barker. “The companies are under sufficient pressure so that within a month or two they’ll all have products — and that [time lag] shouldn’t make too much of a difference.”
Now it’s up to vendors to convince users like Rockefeller’s Khitrov that their whole-genome chip will deliver sound science at a reasonable price. Says Khitrov: “I’d like to see something easier to use, and more easily accessible. And by that I mean price.”
Whole Genome Options: How They Stack Up