Microarray Milestones

The late 1980s were heady days for molecular biologists. PCR was only a few years old. The automated DNA sequencer had just been invented. The Human Genome Project was being debated in Congress. Against such a backdrop, two events on either side of the Atlantic Ocean hardly registered at all.

At Oxford University Edwin Southern, of Southern blot fame, developed a way to use inkjet printing – using four bases rather than four colors – to build oligonucleotide sequences on glass slides like Legos on a solid surface, one atop another. At the same time a young biochemist named Stephen Fodor at a little-known Palo Alto, Calif.-based company called Affymax, began tinkering with using photolithography – a technique used to etch the tiny features of semiconductor chips – to do the same thing.

Aware that they had something powerful by the tail, Southern and Fodor toiled in...


By 1991 Fodor and his colleagues were able to build arrays of peptides and dinucleotides, and they described their success in a proof-of-concept paper that made the cover of Science.1 Shortly thereafter, Fodor left Affymax to found a new startup, Affymetrix, which would soon begin to manufacture commercially what Fodor described in 1993 as "miniaturized peptide and oligonucleotide arrays."2 Southern went on to found Oxford Gene Technology to commercialize his inkjet-based approach.

Affymetrix's first arrays contained sequences just eight nucleotides in length, less than a third of the 25-mer oligonucleotides Affymetrix uses today. Octamers are too short to target genes with a high degree of specificity, but Affymetrix researchers persevered, seeing tremendous promise in the basic math of combinatorial chemistry: While the number of possible DNA sequences 25 bases long is a daunting 4 to the 25th power, the number of chemical steps it takes to create all those sequences is only 4 times 25.

"As you get a linear decrease in size, you get a squared increase in content, and with linear steps in chemistry, you get exponential access to diversity. Those things together are just killers," says Fodor. "We decided we would prepare ourselves and our technology for the day when the Human Genome Project was complete, so we would have the technology to print the human genome."

But not everyone shared his enthusiasm. "There were an awful lot of application-based folks in industry that didn't see the vision of where we were going," he recalls. "They'd go to diagnostics people and say, 'These guys over at Affymetrix are trying to make arrays with a million different features on them, what are we going to do with them?' And they'd say, 'I don't know, we can do everything we need with 10 probes."'


Meanwhile, in another part of Silicon Valley, Stanford University biochemist Patrick Brown was seeing spots. "I had a mental image of a DNA microarray, even including the red and green fluorescent spots, a few years before I'd figured out the details of making them," he writes in an E-mail.

Brown's idea was for an array fabrication method fundamentally distinct from both Fodor's and Southern's: Instead of building oligonucleotides on the array in situ, Brown proposed arranging already-synthesized snippets of DNA at regular intervals on a surface. "I actually proposed the idea ... in an NIH grant proposal that was resoundingly rejected in 1992. It got the worst priority score I'd ever seen. In fact, the study section essentially suggested that if I removed the specific aim related to development of DNA microarrays, and a couple of other forward-looking aims, they would fund it."

Undeterred, Brown continued to pursue microarray development, which he thought would vastly accelerate his work with genomic mismatch scanning. He envisioned a robot that would deposit very small volumes of DNA at regular intervals on a glass surface, and recruited graduate student Dari Shalon to build the prototype.

"The earliest robot we built was made with very cheap parts, and the printing tips were made from the tips of EM [electron microscopy] tweezers held together with epoxy," Brown writes. "The idea came from having many times had the unfortunate experience of these tweezers wick up staining solutions when trying to pick up an EM grid." The resulting arrays used PCR-amplified DNA, and had about 6,000 spots each.

Around the same time, Mark Schena, then a graduate student in Ron Davis's lab at Stanford, was hitting a wall with his work on Arabidopsis homeobox genes. Schena saw in Brown's spotted microarrays a chance to take a far broader look at gene expression than was possible with traditional techniques; Brown saw in Schena's work the kind of biological problem that would prove his technology in action.

"What became clear was that the technologies available to study gene expression were sufficiently below what was really needed to study those genes in detail," says Schena.

A collaboration between the two labs yielded a groundbreaking 1995 Science paper that used the word "microarray" for the first time, and put the spotted microarray on the map.3 Two years later, the Brown lab would publish the first whole-genome microarray study of gene expression, beating out commercial manufacturers to be the first to put the recently completed yeast genome on a chip.4

Eschewing the temptation to commercialize its success, Brown's lab posted instructions for building a homemade spot-printing robot on its Web site. Before long, the do-it-yourself microarray movement had taken hold in many academic labs, where researchers wanted custom slides without the high cost of commercial arrays, and were willing to tinker to get results.

These days commercial chips have become inexpensive enough, and their quality, reliability, and variety good enough, that labs just breaking into microarrays might be better served by commercial rather than by homegrown chips. Still, spot printing remains a relatively cheap and flexible option for many labs and core facilities.


Surprisingly, neither Affymetrix's in situ manufactured chips nor its competitors' spot-printed arrays has been declared a clear winner in terms of accuracy and reproducibility of results. Several studies, most recently a series in Nature Methods,5 have found that both approaches are prone to error and misinterpretation, and both can deliver good results in experienced hands.

Slow Road to the Clinic

Late last year, Roche Diagnostics and Affymetrix teamed up to release the AmpliChip CYP450. Designed to screen patients for genetic variations in drug metabolism, the array is the first approved by the Food and Drug Administration for use as a diagnostic. Other such products are in the pipeline, as are gene-expression based arrays, which have been used in research labs to stratify cancers and predict clinical outcome.

But a lack of standardized protocols for experimentation and analysis is stymying progress. "From a variability point of view, the actual array manufacturing is exquisitely robust. Where most of the noise comes in is, how do you collect your sample, how do you prepare your sample, and how do you do your experiment?" says Affymetrix CEO Stephen Fodor. "The next thing is regulatory-ready standardization in protocols, to be able to have something that's useful in the clinical setting." Complicating the picture are the many different algorithms that can be used to process microarray data before final analysis.

Several efforts are underway to fill the standards gap. The Microarray Quality Control (MAQC) Project, for instance, is collecting gene-expression data (both microarray and quantitative RT-PCR) from two human reference RNA samples to establish benchmarks against which new platforms, procedures, or algorithms can be tested. Labs can then demonstrate their technical proficiency by running these RNAs in their labs and checking their data against MAQC values. Another project, the External RNA Controls Consortium, is developing a set of "spike-in" RNAs that can be added directly to experimental RNA samples as external controls to monitor the performance of microarray processes.

Though such controls and benchmarks may ultimately be required of labs that wish to run gene-expression analyses for diagnostic purposes, or for companies submitting array data to the FDA, they remain, for the moment, research projects, says Leming Shi, MAQC coordinator. "We would like to see some indication that the pharmacogenomics labs can demonstrate their proficiency and quality and that the data are analyzed appropriately", he says. "But we don't know yet if this will be required or not to make a regulatory decision."

Meanwhile, other companies, like Agilent Technologies and Oxford Gene Technologies, have refined and capitalized on inkjet printing methods, bolstered by interest from giants like Hewlett-Packard and Merck. Still others, such as CombiMatrix, have harnessed developments in microfluidics to find new ways of generating microarrays.

Nevertheless Affymetrix retains the lion's share of the microarray market – and it fights hard to keep it. Beginning with a pitched battle over the Southern patents, the company has been embroiled in a string of intellectual property lawsuits and disputes over licensing agreements. One source of perpetual consternation to its competitors is the fact that Affymetrix claims its patents cover not only the photolithographic technique, but also the high density of its arrays as well.

To a large extent, Affymetrix's position on that front is secured by both patents and physics, as an array's density is a function of the technology used to create it. Inkjet and spotting methods are constrained to a maximum of about 50,000 features on a slide by the minimum size of a droplet, whereas features constructed with semiconductor technology are constrained only by the size of a manipulable beam of light. The most recent generation of Affymetrix chips has six million features, each five microns across, and the company has prototype chips with one-micron features.


With such vast numbers of features per chip, researchers are free to consider studies never before possible. Scientists armed with "genomes-on-a-chip" can now easily scan the expression of hundreds of thousands of genes for the few that might be involved in a form of cancer or a genetic disorder. Using San Diego-based Illumina's latest BeadChip arrays users can test expression of more than 46,000 human genes, in each of six samples simultaneously, with 30-fold redundancy per feature. Researchers can also screen individuals for thousands of genetic polymorphisms called SNPs (single nucleotide polymorphisms). One company, Perlegen Sciences (an Affymetrix spin-off), has the ability to screen 1.5 million human SNPs at once.

Though such "fishing expeditions" look impressive, and get a lot of press, they represent only part of the microarray's promise. For one thing, the technology has morphed to include new kinds of arrays, including tissue arrays (a hybrid of microarrays with traditional slide-based pathology), cell arrays, and most promisingly, protein arrays.

Fast Facts

How has the microarray transformed the life sciences: Induced a shift from hypothesis-driven to discovery-driven research

When was it developed: 1988; first application 1995

Primary application: Gene expression analysis and genotyping

Pros: Survey an entire genome on one slide

Cons: Data analysis and reproducibility remain perennial problems

Key references: S.P. Fodor et al., "Light-directed, spatially addressable parallel chemical synthesis," Science, 251:767–73, 1991; M. Schena et al., "Quantitative monitoring of gene expression patterns with a complementary DNA microarray," Science, 270:467–70, 1995.

Clinical application: Personalizes medicine, whether via molecular stratification of disease or through pharmacogenetics

DNA arrays, too, are morphing. "A lot of the microarray work, if you go back to the early days of microarray analysis, they looked for genes that were varying, annotated them, and that was enough to get a publication," says University of Colorado researcher Ron Taylor. "Nowadays you need to look for biological relationships, causation. Once you know the subset of genes that is of interest in a particular type of cancer, you would probably want to decrease the number of genes on the chip."

Clinical applications will likely require more focused chips with fewer features, if only for ease of analysis. The first FDA-approved diagnostic microarray, the Amplichip CYP450, a collaborative venture between Affymetrix and Roche Pharmaceuticals released last year, screens individuals for polymorphisms at just two genes, variants of which can influence how patients respond to certain medications.

But with researchers constantly rolling out new applications the high-density microarray is not likely to outlive its usefulness anytime soon. "The cool thing that is going on now ... is the application of arrays to environmental samples: metagenomics, or environmental genomics, or community genomics," says Joe DeRisi, a former student of Brown's who is now at the University of California, San Francisco. DeRisi made headlines in 2003 when he used a microarray to identify the SARS virus.6

The most recent generation of high-density chips sport feature densities large enough to "tile" the entire length of the human genome at five-nucleotide intervals, making it possible to screen for gene expression in areas long thought to be "junk DNA." Such arrays are now showing activity in some unexpected areas of the genome, says Fodor. "If you ever say "junk DNA" to Paul Berg ..." he laughs. "You have to have some confidence that the human genome is that large for a reason."

Microarray Milestones


Edwin Southern files UK patent applications for in situ synthesized, oligo-nucleotide microarrays


Stephen Fodor and colleagues publish photolithographic array fabrication method


Undeterred by NIH naysayers, Patrick Brown develops spotted arrays


Affymax begets Affymetrix


Mark Schena publishes first use of microarrays for gene expression analysis

Edwin Southern founds Oxford Gene Technologies


First human gene expression microarray study published

Affymetrix releases its first catalog GeneChip microarray, for HIV, in April


Stanford researchers publish the first whole-genome microarray study, of yeast


Brown's lab develops CLUSTER, a statistical tool for microarray data analysis; red and green "thermal plots" start popping up everywhere


Todd Golub and colleagues use microarrays to classify cancers, sparking widespread interest in clinical applications


Affymetrix spins off Perlegen, to sequence multiple human genomes and identify genetic variation using arrays


The Microarray Gene Expression Data Society develops MIAME standard for the collection and reporting of microarray data


Joseph DeRisi uses a microarray to identify the SARS virus

Affymetrix, Applied Biosystems, and Agilent Technologies individually array human genome on a single chip


Roche releases Amplichip CYP450, the first FDA-approved microarray for diagnostic purposes

Interested in reading more?

Magaizne Cover

Become a Member of

Receive full access to digital editions of The Scientist, as well as TS Digest, feature stories, more than 35 years of archives, and much more!
Already a member?