The companies in this article were selected at random. Their inclusion in this article does not indicate endorsement by either AAAS or Science, nor is it meant to imply that their products or services are superior to those of other companies.

This is the last of four supplements this year on biochips. The first three appeared in the 27 February, 14 May, and 9 July issues of Science.

At first glance, a DNA microarray appears rather ordinary, consisting as it does of a series of oligonucleotides on a substrate. But in just the past few years this simple item has morphed from a rather obscure phenomenon explored by academic specialists into a tool that has wide application throughout the world of life science. “Just as every lab now has a PCR facility, every lab will soon have easy access to microarray systems,” says Gustavo Stolovitzky, head of functional genomics and systems biology center at IBM Research.

For biological researchers, the use of DNA microarrays represents a way of building on the advances in understanding created by genome sequencing. “It’s the core technology that a lot of genomics revolves around,” says Tony Gordon, genomic diagnosis business manager for the British unit of German firm MWG Biotech. “With DNA microarrays you’re capable of looking at the entire genome expression profile in one experiment,” adds Mindy Lee, director of marketing and product management for the CodeLink microarray platform at GE Healthcare (formerly Amersham Biosciences).

Even though it is continuing to evolve, the technology has found significant applications. “The DNA microarray is still not a fully mature technology; a lot of parameters are being optimized,” says Sean Yu, vice president of operations for SuperArray. “But the scientific community at first and then the industrial community have accepted it as a powerful tool to analyze genes.”

Biopharmaceutical firms in particular have taken up the technology. “There’s no question that DNA microarrays are becoming increasingly important tools for drug discovery,” says Reid Huber, associate director of Incyte Corporation. “The use of technology has traditionally focused on early-stage research. But as our sophistication in experimental design and data analysis has matured over the past few years, applications of the technology have blossomed. It has become a fundamental component of many aspects of drug discovery research.” Doug Bassett, head of informatics at Rosetta Inpharmatics, a wholly owned subsidiary of pharmaceutical firm Merck, agrees. “Microarrays,” he says, “have really evolved into a core technology supporting therapeutic discovery efforts.”

Merely a Beginning

That evolution represents merely a beginning. “As a platform for finding targets for intervention, microarrays haven’t reached their full potential yet,” says Todd Martinsky, founder and executive vice president of TeleChem International. “But in the long term they will be used in predictive, preventive, and personalized medicine.” In fact that has already started. “Scientists are beginning to correlate patients’ profiles with their prognoses,” says Siobhan Pickett, director of genomic systems for Molecular Devices (formerly Axon Instruments). “Several groups are starting experimentally to offer cancer treatment options based on genetic profiles.”

The technology’s success has created one problem that causes serious ambivalence among many life science researchers. “You have a huge amount of data you need to traverse statistically to make sense of it,” says Vivian Bonazzi, chief scientist in the informatics division of Invitrogen. That quantity “is the main advantage and the main disadvantage of microarrays,” MWG Biotech’s Gordon adds. “An average core facility that does, say, 1,000 or 1,500 slides annually might create a terabyte of data in a year. If you don’t deal with that precisely, you’ll get into hideous difficulty.”

Overcoming that difficulty and making sense of microarray data demand a combination of sophisticated hardware and bioinformatics software — a combination that can turn many bench scientists into jelly. After all, Bonazzi points out, “Biologists fall into two categories: Those who come out in some sort of rash when they see a computer and those who don’t.”

The former group needs help and reassurance that vendors willingly provide. “We rejoice in data,” declares IBM’s Stolovitzky. “We interact a lot with biological laboratories and discuss what they expect. We devise methodologies for digging out the data that they want.” Rosetta takes a similar approach. “We create interdisciplinary teams of life scientists working with high caliber data analysts, including particle physicists and signal processing experts who discover and leverage new approaches and algorithms for analyzing the data,” Bassett explains. “When we find a new method that works, we transition it into our software platform.”

Beyond Software

Dealing with the deluge of data requires more than analytical software. “For DNA microarrays, the right oligo design on the chip can make the analysis much more straightforward,” TeleChem’s Martinsky explains. Adds Huber of Incyte: “The real power of the technology comes when one can analyze many diverse data sets together.”

Life scientists don’t have to rely on their own skills to handle the complexities of microarraying. Vendors of tools and technologies play significant roles in facilitating the design and analysis of microarray experiments in user friendly ways that permit scientists to spend their time thinking about their research rather than dealing with the mechanical quirks of the technology. “One barrier that prevents a lot of scientists using microarrays is the technical complexity of everything from target labeling to image acquisition,” says SuperArray’s Yu. “To mitigate that problem, vendors have worked diligently to provide systems with automatic handling and other benefits.” Adds Gordon of MWG Biotech: “Vendors offer wares that address solutions to the data problem. The community that makes freeware also helps. The vendors also have some very nice LIMS [laboratory information management systems] software to control a lot of the functions of data analysis.”

Suppliers have also begun to use the Internet to enable data analysis. “One trend I’m particularly fond of is analysis software available over a web browser, such as Gene Sifter, so that users are not required to upkeep powerful computer systems and software versions on site,” TeleChem’s Martinsky explains. “By doing analysis through the Internet, you can naturally take advantage of information in the public domain and get the most recent databank information,” adds Yu, whose company will soon release software capable of combining image analysis and data analysis via the web.

Huber of Incyte gives a user’s view of microarray analysis software. “We have chosen to license in externally developed solutions,” he says. “They enable us to direct our energies to applying our platform to our internal needs.” The company has a particular relationship with Rosetta Biosoftware. “Its Resolver platform provides a foundation for our internal efforts and has contributed to the successful application of the technology to several of our drug discovery programs,” Huber continues.

Biology Versus Brownie Points

Scientists who want to work with DNA microarrays don’t have to start from scratch. “Increasingly, microarrays are going the way of genome sequencing — becoming just another technique,” says Gordon. “Scientists are interested in the biology rather than getting brownie points for doing the technique. So why bother to learn how to do it when you can have us do it for you?”

Several vendors offer microarray design facilities for scientists new to the technology or too busy to worry about the mechanical details of their studies. “We offer microarray design services for gene expression, SNP, and splice variant analysis,” says Martinsky. “We have highly trained people who work closely with the researchers and the biological goals of the research project to get the microarray designed right. We are not simply running sequences through computers; that’s not enough to get the job done.”

SuperArray also works closely with clients to customize their microarrays. “We have an internal array design team and we collaborate with academic researchers to see specific needs,” Yu says. “When we design our arrays, we cater to specific applications. That limits the scope of the array. It includes 100 or 200 genes instead of 10,000, thus greatly simplifying the data analysis requirements.” The company has developed specific expertise in microarrays for research on stem cells, and offers mouse stem cell and human stem cell arrays. “We cater our arrays specifically for stem cell research with potential applications down the road in the clinic,” Yu continues.

Vendors also supply scientists who want to design and set up their own microarrays. “CodeLink is our microarray platform,” GE Healthcare’s Lee says. “We also have reagent kits and labeling kits to help scientists deal with their data. And we have a software package that allows customers to extract quality control reports and spot details so that you know which gene goes with which spot on the microarray.”

Progression of Data Gathering

Once scientists have designed their microarrays and completed their experiments with them, they face the critical step of reading the data they have generated. The most common methods rely on fluorescent systems that perform several basic operations for microarrays: They excite fluors attached to the samples, collect emitted light, and generate digital images of the fluorescent signal.

To read their microarrays, scientists usually opt for one of the two most common methods of detection: imagers and scanners. Imagers use filtered white light that they detect with a charge coupled detector (CCD) while scanners rely on laser excitation and a photomultiplier tube (PMT) detector. “A CCD camera takes a snapshot of a specified area on a slide,” explains Pickett of Molecular Devices. “In a laser based system, the light scans across the slide.” What factors control a scientist’s decision on the technology to use? “The biggest issue in deciding on scanners or imagers is the quality of the output,” says Lee of Amersham. “We provide our users with application notes so that they can get the best quality data possible.”

Imagers start with white light from a xenon or mercury lamp. They control the excitation wavelength by filtering out all but a narrow range of wavelengths. The lamp illuminates a relatively large area of the microarray slide, and a stationary CCD detector collects the fluorescent light emitted from the entire field of view. The system then converts the signal intensity at each pixel location into a digital image. Because no CCD imager has a large enough field of view to take a snapshot of an entire microarray slide, detection requires a series of shots that are then stitched together to make a comprehensive image of the slide.

The Scanning Alternative

For the scanning alternative, as for CCD imaging, scientists use fluorescent labels from such companies as Amersham Biosciences and Molecular Probes (an Invitrogen company) to tag the DNA on their slides and detect them with laser scanners specifically designed for use with microarrays. The scanners often include software for analyzing and interpreting the data. Scanners can use confocal technology to eliminate unwanted background fluorescence by limiting the range for picking up signals to distances above the plane of the array where the substrate is located. However, some scientists feel that confocal technology is unnecessary for microarray scanning, because the primary source of background is nonspecific hybridization in the same focal plane as the sample of interest.

Scanners use one or more lasers to scan samples or features on a microarray surface; scanning at several wavelengths permits scientists to detect multiple fluors. The laser beam scans back and forth across a sample, exciting one pixel at a time. The light emitted from the excited fluor travels back through the excitation lens and is collected by the PMT. The PMT amplifies the signal, which is then converted into a digital value that contributes to an image of the signal’s intensity at each pixel location. Companies that offer imagers of these types include Affymetrix, Agilent Technologies, Hitachi Genetic Systems, PerkinElmer Life and Analytical Sciences, and many others.

Affymetrix offers scanners and integrated software. The company’s GeneChip Scanner 3000 incorporates advanced designs with smaller size, improved performance, and lower than usual variation from one scanner to another. The device provides more accurate gridding and more consistent scanner-to-scanner biological performance, improving data integrity and data sharing between researchers. The company sells it with a powerful computer work station, loaded with GeneChip Operating Software.

The GenePix 4000B microarray scanner from Axon Instruments can scan two wavelengths simultaneously, thereby reducing scanning time. It features user-selectable laser power and focus position. It also includes both GenePix Pro software for image acquisition and first-pass analysis and Acuity software, a scalable database, statistical, and data visualization package. “We have focused our attention on imaging accuracy,” Pickett says. “Each of our scanner models meets the same set of performance criteria for the imaging specifications. They also feature ease of use.”

Another method of microarray detection involves radioactive labels that can be imaged with a phosphorimager or — less glamorous but still effective — autoradiography film. MP Biomedicals and PerkinElmer offer phosphorus isotope radiolabels for this procedure. Scientists most often use the radiolabels with nylon membrane macroarrays, such as those offered by BD Biosciences Clontech and Millipore.

Interpreting the Results

Once they have scanned their arrays, scientists need to interpret the results. Because individual microarrays can contain thousands of samples or spots, they can produce huge volumes of data. Storing and analyzing the information can cause a serious bottleneck in laboratory research. To avoid the complexity of working with such vast amounts of data, some researchers perform array experiments first with large comprehensive chips, such as Affymetrix’s gene on a chip products, and then down-size their research efforts by focusing on a specific family of genes.

Certainly, making sense of the large quantities of information collected from gene sequencing and gene expression experiments is no simple task. Scientists can easily spend many hours each week working with computers and specialized software to store and manage sequence data, design microarray formats, and analyze the data gathered from these studies. To minimize that nonresearch time, companies such as Genetix, Invitrogen, Molecular Devices, and TeleChem offer software and other products for microarray analysis. “The more you can free up researchers to be creative rather than sitting in front of the scanner,” says Pickett of Molecular Devices, “the faster they will be able to progress in their work.”

Invitrogen offers Vector NTI Advance and Vector Xpression 3.1 software packages for microarray data analysis. The company has developed a broad range of software modules that are fully integrated to allow researchers to move easily from one application to another. “We’re trying to integrate known pathways and interactions with DNA results,” explains Paul Predki, Invitrogen’s head of R&D. “Doing that enables us to map pathways in a single day.” Adds Hollis Kleinert, president and CEO of Invitrogen subsidiary Protometrix: “Our system helps to drive the results. The system helps give us the information and the knowledge of what to do next.”

Bioinformatic Solutions

Rosetta Biosoftware offers bioinformatics software solutions for drug discovery and the development of pharmaceutical and agricultural products. The Resolver Gene Expression Analysis System, the company’s flagship product, includes tools and features for high-powered analysis, visualization and storage of gene expression data. The Rosetta Luminator system provides an affordable, turnkey enterprise solution with the advanced analysis features and performance needed by small to mid-size commercial and academic gene expression research efforts. “One of our biggest applications of this software is biomarker discovery,” says Bassett of Rosetta Inpharmatics. “We also use it for new target identification and validation.”

Vendors have also developed software tools to help scientists manage those sets of data, derived in different formats from different experiments on different platforms. Silicon Genetics and Spotfire offer software solutions for scientists who need to manage and analyze the results from research with DNA microarrays. MWG Biotech offers a one-stop shop type of service. “We work with Tecan to scan customers’ slides, Gordon says. “Beyond that we do the image and data analysis and put it in LIMS.”

For scientists who prefer to do it themselves, The Institute for Genomic Research (TIGR) has developed a package of publicly available software programs for analyzing microarrays. Its MIDAS (MIcroarray Data Analysis System) permits scientists to analyze raw experimental data through a pipeline that they design themselves. The MADAM (for MicroArray DAta Manager) application gives users the ability to load microarray data into and retrieve it from a database. The TIGR MultiExperiment Viewer enables scientists to identify patterns of gene expression and differentially expressed genes in microarray data. And the Spotfinder is a software tool designed for image processing using the TIFF images generated by most microarray scanners.

Computing and Collaboration

Applications of microarrays are clearly pressing the limits of conventional computing capacity. To increase computing capabilities, IBM, Hewlett-Packard, Apple, and other vendors have set out to gain more power from existing computers and to develop more powerful computers and more capable software.

IBM Life Sciences has developed a microarray storage and retrieval solution that its scientists have used in collaborations with Johns Hopkins University and The Mayo Clinic. This solution complements Big Blue’s Genes@Work, a system for analyzing gene expression data derived from DNA microarrays. The system provides visualization tools, machine learning and sample classification modules and a computationally intensive pattern discovery algorithm. Scientists can use such tools to find patterns of gene expression that differentiate between populations of case and control samples, such as cancerous versus normal cells.

In addition to improved software, says IBM’s Stolovitzky, “We realize more and more the need to have higher power computing.” One approach to boosting computational horsepower is the use of grid computers. This solution uses the power of interconnected numbers of computers to run demanding applications. A similar approach involves hooking up large numbers of individual microprocessors to deliver faster computing times. IBM is also collaborating with the Translational Genomics Research Institute to develop Boolean networks that will help to analyze microarray data, among other tasks.

Collaboration among vendors or between suppliers and academic departments represents a growing trend. Affymetrix, for example, is partnering with Caliper Life Sciences to expand its offering from microarrays to microfluidics.

Setting the Standards

Collaboration plays a particularly strong role in the effort to develop standards for microarrays. “Standards are extremely important,” says Stolovitzky. “Otherwise we are in a Tower of Babel in which we all talk in different languages.” Leading the way in developing microarray standards are United States government agencies such as the National Institute of Standards and Technology and the Food and Drug Administration. The Microarray Gene Expression Data Society, an international organization of biologists, computer scientists, and data analysts who want to broaden the sharing of microarray data, is also involved in setting standards. In the commercial world, Affymetrix has worked with several organizations to develop a set of standards for microarrays. In addition, says Lee of Amersham, “We’re in discussions to help develop a standard for microarray data submitted for publication.”

DNA microarrays have plainly proven their utility as valuable experimental platforms for research. As progress continues in developing standards for these tools, and as more application-specific arrays come to market, microarrays will become increasingly effective in revealing useful scientific data.

Peter Gwynne (//pgwynne767{at}aol.com) is a freelance science writer based on Cape Cod, Massachusetts, U.S.A. Gary Heebner (//gheebner{at}cell-associates.com) is a marketing consultant with Cell Associates in St. Louis, Missouri, U.S.A.