Biotechnology Advances 18 (2000) 35–46

0734-9750/00/$ – see front matter © 2000 Published by Elsevier Science Inc. All rights reserved.PII: S0734-9750(99)00035-X

Research review paper

cDNA microarray technology and its applications

Charlie C. Xiang

a,

*

,1

, Yidong Chen

b,2

a

Department of Molecular Virology, Virology Division, US Army Medical Research Institute of Infectious Diseases, Ft. Detrick, Frederick, MD 21702, USA

b

National Human Genome Research Institute, National Institutes of Health, Bethesda, MD 20892, USA

Abstract

The cDNA microarray is the most powerful tool for studying gene expression in many different or-ganisms. It has been successfully applied to the simultaneous expression of many thousands of genesand to large-scale gene discovery, as well as polymorphism screening and mapping of genomic DNAclones. It is a high throughput, highly parallel RNA expression assay technique that permits quantita-tive analysis of RNAs transcribed from both known and unknown genes. This technique provides di-agnostic fingerprints by comparing gene expression patterns in normal and pathological cells, and be-cause it can simultaneously track expression levels of many genes, it provides a source of operationalcontext for inference and predication about complex cell control systems. This review describes thisrecently developed cDNA microarray technology and its application to gene discovery and expression,and to diagnostics for certain diseases. © 2000 Published by Elsevier Science Inc. All rights reserved.

Keywords:

cDNA microarray; Technology; RNA expression assay technique

1. Introduction

To transform a single fertilized egg cell into an adult human body and then keep that bodyalive and healthy requires about 100 000 genes, each of which must adjust its expression toprecise degrees and at precise times and locations. Three billion base pairs of the human ge-nome will be completely sequenced by the year 2003. The next great challenge is to find outhow those genes act in concert to regulate the whole organism. To understand the develop-ment of organisms, the onset of genetic diseases, or the concerted functions of genes in regu-

* Corresponding author. Tel.:

1

1-310-267-1947; fax:

1

1-310-794-5546.

E-mail address

: cxiang@ucla.edu (C.C. Xiang)

1

Current address: Microarray Core Facility, University of California, Los Angeles, Room 5554, Gonda Cen-ter, 695 Young Drive South, Los Angeles, CA 90095.

2

NIH/NHGRI/CGB, Building 49, Room 4B24, 49 Convent Drive, MSC 4470, Bethesda, MD 20892-4470.Tel.:

1

1-301-402-3150; fax:

1

1-301-402-3241.

E-mail address

: yidong@nhgri.nih.gov

36

C.C. Xiang, Y. Chen / Biotechnology Advances 18 (2000) 35–46

lating cell function or transformation, the best procedure is to monitor the fluctuating activi-ties of the genes in different tissues at different stages of development, in good and in badhealth. However, the complete mechanism or correlation among the genes has not yet beenelicited for even a single cellular function. Monitoring the activities of a panel of genes to de-termine the role of genes in regulating any biological process in a whole organism is a formi-dable job. Until 3 years ago it was only possible to tackle one gene at a time. The DNA mi-croarray technique offers the best way to approach such a daunting task. This technology hasbeen steadily developed for more than 3 years since Patrick Brown and his colleagues firstpublished their work in October 1995 [1]. It seems likely to become a standard tool of bothmolecular biology research and clinical diagnostics.

There are basically two types of microarrays that have been developed so far: cDNA mi-croarray and oligonucleotide array. Arraying methods include on-chip photolithographicsynthesis of 20–25 mer oligos onto silicon wafers [2], off-set printing of 20–25 mer oligosarray, and 500–5000 bp cDNAs printed onto either glass slides or membranes [3]. An idealsupport allows effective immobilization of a target onto its surface, and robust hybridizationof probe with the target. Glass has many of the same advantages as nylon, which is the otherstandard support used for making microarrays. It also has unique advantages including cova-lent attachment of DNA samples onto a treated glass surface; durable material that sustainshigh temperatures and washes of high ionic strength; nonporous support so the hybridizationvolume can be kept to a minimum, thus enhancing the kinetics of annealing probes to targets;no significant contribution to background noise because of its low fluorescence; and hybrid-ization of the array with two or more probes labeled with different fluors for serial or parallelanalyses [4]. This review will focus on cDNA microarrays on glass and its applications.

2. Biological aspects of cDNA microarray technology

Life depends on the ability of cells to store, retrieve, and translate the genetic instructionsrequired to make and maintain a living organism. Hereditary information is passed on from acell to its daughter cells at cell division, and from generation to generation of organismsthrough the reproductive cells. These instructions are stored within every living cell as thegenes—the information-containing elements that determine the characteristics of a species asa whole and of the individuals within it. A cell relies on its gene products for a wide varietyof functions including energy production, macromolecule biosynthesis, cellular architecturemaintenance, and response to environmental stimuli. Proteins are the active working compo-nents of the cellular machinery, whereas DNA stores the information for protein synthesisand RNA carries out the instructions encoded in DNA. Expression of the information inDNA is mediated by RNA. Messenger RNAs (mRNAs) are the transcripts that carry the spe-cific information for the sequence of amino acids in proteins. There are very many mRNAspecies of widely varying sizes, generally possessing little secondary structure. With rare ex-ception, all species of eukaryotic mRNAs are polyadenylated. The amount of mRNA inmammalian cells has been estimated at approximately 500 000 mRNA molecules per cell.They are relatively stable in eukaryotes. The half-life of mRNAs in mammalian cells can lastfor days. Generally speaking, the level of a mRNA in a particular cell represents the meta-

C.C. Xiang, Y. Chen / Biotechnology Advances 18 (2000) 35–46

37

bolic activity of the specific gene. The human cell contains about 50 000–100 000 genes.Some genes are expressed in all cell types. Others are only expressed in particular cell types.As cells change their status during their development or the cells are stimulated by physicaland chemical reagents, the gene expression patterns in the cells will be altered. For example,the optimal intracellular environment for viral replication is different from the optimal envi-ronment for normal cell growth and replication during viral infection. The virus itself hasevolved the means to alter the intracellular environment to suit its own ends. Some of thesealterations in the intracellular environment result from the changes in the concentrations ofcellular RNAs. Comparison of gene expression patterns by measurement of mRNA levels innormal and pathological cells could provide useful diagnostic ‘fingerprints’ and help identifyaberrant functions that would be reasonable targets for therapeutic intervention.

3. Bioinformatics in cDNA microarray technology

During the early stages of microarray technology, much of the effort went into the develop-ment of the enabling technology, such as serial analysis of gene expression (SAGE) [5,6], oli-gonucleotide arrays, and cDNA arrays. At the same time, relatively little attention had beenpaid to data analysis methods and interpretation of gene expression. One of the reasons is dueto data insufficiency of the gene expression pattern. The same situation was endured in theearly stage of DNA sequence database development. Only after more than 15 years of accumu-lation of DNA sequences in GenBank were many of sequence-based tools like BLAST devel-oped. With the introduction of microarray technology, geneticists face the rapid evolution ofbioinformatics driven not only by the problems associated with gene mapping and sequencing,but also by the massive quantities of data generated by parallel expression monitoring technol-ogy. Currently, with the advances of the genome-scale technologies, the biomedical researchpublications have shifted their attention to the functional aspects of thousands of genes, insteadof a single gene’s structural information. However, the bioinformatics community still facesmany of the challenges posed by the parallel approaches of gene expression study. Some ofmost pressured subjects include: (1) laboratory information management, (2) microarray imageanalysis, (3) gene expression profiling analysis, (4) genetic fingerprint or expression patternclassification, and (5) other high-level control system modeling issues.

3.1. Laboratory information management system

The main goal of laboratory information management system (LIMS) is to track and man-age material information flow [7]. The ArrayDB (http://www.nhgri.nih.gov/DIR/LCG/15K/HTML/) was developed to store, retrieve, and analyze microarray experiment information.For each microarray experiment, the basic information in the database consists of: (1) infor-mation about microarrays and specific cDNA clone inserts (such as clone ID, title, databasehyperlink, etc.); (2) information about probes (such as sample source, experiment condition,fluor-type, etc.); and (3) information about images (raw scanned image, intensities and ratiosfor each target clone, etc.). The other goal of ArrayDB is to identify patterns and relation-ships among intensity ratios, both individual and across multiple experiments. In addition,

38

C.C. Xiang, Y. Chen / Biotechnology Advances 18 (2000) 35–46

the option to download data associated to microarray experiment over the Internet adds flex-ibility to the end-users’ customized data analysis.

3.2. Microarray image analysis

The objective of microarray image analysis is to extract probe intensities or ratios at eachcDNA target location, and then cross-link printed clone information so that biologists caneasily interpret the outcomes and perform further high-level analysis. However, the microar-ray image sources are not only from one print-mode (i.e. different printing tip arrangement ordifferent arrayers) [8] or one hybridization method (i.e. fluorescent [Stanford, NIH, etc], ra-dioactive probe, and others) [9], the analysis methods are very different. Typically, microar-ray image analysis consists of cDNA target segmentation, target detection, local backgroundintensity and probe fluorescent intensity measurement, ratio analysis, and, finally, data visu-alization. The keys to success of a single slide analysis are the measurement of expression ra-tio between samples and the selection of a set of ‘housekeeping’ genes for ratio calibration.

3.3. Gene expression profiles and fingerprinting

Single microarray hybridization processed by microarray image analysis software may besufficient for much high-throughput gene expression screening. For some other experiments,such as the study of development of some biological systems, monitoring the change of tem-poral gene expression patterns becomes very important to study the biological interactionsbetween genes within an organism [10], to understand the mechanism of some orderly be-havior of gene expressions [11], and to look for relatedness among cell types [12].

The first step of data analysis is often referred to as ‘data exploration,’ in which any nonran-dom patterns or structures requiring further explanation are recognized. Clustering is one of thetechniques in which the data of interest are placed into a small number of homogeneous groupsor clusters. To study the orderliness of expression data, the multidimensional scaling (MDS)technique is routinely used for biological sample similarity visualization. For every gene usedin the microarray experiments, K-mean-based algorithms or hierarchical clustering methods(similar to Eisen et al. [13]) are used for gene expression similarity clustering and visualization.

4. Making cDNA microarray

Fig. 1 describes the basic principles of the DNA microarray technology. Microarrays con-taining large DNA fragments, such as cDNAs, are generated by physically depositing smallamounts of each DNA of interest onto known locations on glass surfaces. They begin with theselection of the targets to be printed on the array. Deposition strategies typically produce mi-croarrays consisting of groups of cDNAs amplified by the polymerase chain reaction (PCR)ranging from 0.5 to 2.0 kb. In many cases, the targets are chosen directly from databases suchas GenBank, dbESt, and UniGene. Additionally, full-length cDNAs, collections of partiallysequenced cDNAs (or ESTs), or randomly chosen cDNAs from any library of interest can beused. The DNA fragments were PCR amplified from individual clone using specific primersor universal primers if all the genes were cloned in the universal vector.

C.C. Xiang, Y. Chen / Biotechnology Advances 18 (2000) 35–46

39

The microarrays were fabricated on poly-L-lysine (Sigma, St. Louis, MO, USA) coatedmicroscope slides. The DNA fragments are cross-linked by UV to the matrix. After fixation,residual amines on the slide surface react with succinic anhydride to reduce the positive chargeat the surface.

A robotic arraying machine loaded about 1

m

L of PCR-amplified fragments from correspond-ing wells of 96-well plates and deposited about 5 nL of each sample onto each of 42–110 slides.Several commercial arraying machines are currently used including MicroGrid from BioRo-botics (Cambridge, UK), GMS 417 from Genetic Microsystems (Woburn, MA, USA), Omni-Grid from GeneMachines (San Carlos, CA, USA), and the PixSys PA series from Cartesian Tech-nologies (Irvine, CA, USA). The arrayer made by Beecher Instruments (Silver Spring, MD,USA) in the National Human Genome Research Institute at National Institute of Health inBethesda, Maryland comprises an xyz cantilever type robot holding 16 quill pen-type probes,a removable vacuum chuck for 48 standard microscope slides, a 20 microtiter tray loader/stacker, a wash/dry station, a controling PC, air-handling components, and a cabinet (Fig. 2).The robot moves only the probe holder. Sixteen stainless steel probes, which function as‘quill pens,’ are spaced on 9-mm centerlines to conform to the well spacing of standard 96-

Fig. 1. Outline of the microarray technology. PCR-amplified and purified DNA fragments are printed on theknown locations of the glass slide to make the DNA array. cDNA probes are prepared separately (e.g. from unin-fected and infected cells) through reverse transcription. The probes are then hybridized to the array. The array isscanned by a scanning confocal microscope. The final microarray images are analyzed by various computer pro-grams. R: red color, G: green color, Y: yellow color.

40

C.C. Xiang, Y. Chen / Biotechnology Advances 18 (2000) 35–46

well microtiter trays. The probes are spring loaded to accommodate small differences inprobe length and slide and well positions. The PC controls all of the other components andallows operator input of various parameters, such as the number of probes, trays, and slides,spot spacing, and pattern on the slides, duration of each cycle component and speeds and ac-celerations of the robot. The arrayer is able to put down 16 spots on each of 48 slides and towash and dry the probes for the next set of cDNAs for the next tray in about 70 s. Most ofthis time is taken up with the actual spotting, as the wash and dry cycles are about 2 s eachand the loading is about 10 s. Thus the contents of one 96-well tray can be spotted every 7min, and 10 000 spots should take about 12 h. Many arraying machines use the ‘quill’-typespotting tips that were originally designed by Pat Brown and his colleagues at Stanford Uni-versity. But other companies developed solid pins and the GMS 417 from Genetic Microsys-tems uses a special device, so-called Pin-Ring (Fig. 3).

5. Probe labeling and array hybridization

Total RNA is typically isolated from different tissues and cell lines using the Trizol re-agent from BRL according to the manufacturer’s instructions. The purity of RNA is a crit-

Fig. 2. Microarray arrayer in the Laboratory of Cancer Genetics, NHGRI, NIH. The arrayer comprises (1) a con-troling PC, air-handling components, and a cabinet; (2) a plate for 48 standard microscope slides, a microtiterplate loader/stacker, a wash/dry station; and (3) an xyz robot holding 16 printing pens.

C.C. Xiang, Y. Chen / Biotechnology Advances 18 (2000) 35–46

41

ical factor in hybridization performance, particularly when using fluorescence, since cellularprotein, lipid, and carbohydrate can mediate significant nonspecific binding of fluorescentlylabeled cDNAs to slide surfaces. Several total RNA extraction kits have been tested toproduce satisfactory results, including Trizol reagents from BRL and RNeasy kits fromQiagen.

Fluorescently tagged cDNA probes from the pair of samples to be analyzed were pooledtogether. The block reagents, such as yeast tRNA, poly dA, and human Cot-1 DNA, wereadded. The final hybridization solution was 3

3

SSC and 0.1% sodium dodecyl sulfate(SDS). The probes were hybridized to an array at 65

8

C for 16–24 h. Unbound probes wereremoved by washing in SSC and SDS solutions at room temperature. The slides were centri-fuged to remove residual liquid and were air-dried. For the detailed protocols for probe label-ing and array hybridization, please refer to the web site of Cancer Genetics Branch in the Na-tional Human Genomic Research Institute at http://www.nhgri.nih.gov/DIR/LCG/15K/HTML.

6. Array scanning and data analysis

To determine which DNAs correlate with changes in gene expression or toxic effects, themicroarrays are first scanned to produce visual images and to generate raw numerical datafor each spot on the array. The microarray reader is basically a computer-controlled invertedscanning fluorescent confocal microscope with a double or multiple laser illumination sys-tem, such as ScanArrayer 4000 and 5000 from General Scanning (Watertown, MA, USA),Avalanche from Molecular Dynamics (Sunnyvale, CA, USA), GMS 418 from Genetic Mi-croSystems (Woburn, MA, USA), and GeneTAC from Genomic Solutions (Ann Arbor, MI,USA). In addition, some companies, such as Genometrix (The Woodlands, TX, USA; Ap-plied Precision, Seattle, WA, USA), are developing charge-coupled device (CCD) camerasto capture the microarray images. The illumination system of the NHGRI Scanner manufac-tured by Beecher Instruments (Silver Spring, MD, USA) consists of three air-cooled lasers

Fig. 3. Pin-and-ring array technology by Genetic MicroSystems. The pin-and-ring array technology is capable ofcreating spots of extremely consistent size, shape, and volume (Courtesy of Genetic MicroSystem.)

42

C.C. Xiang, Y. Chen / Biotechnology Advances 18 (2000) 35–46

(Fig. 4): a 488-nm, 100-mw Argon ion laser for exciting FITC; a 532-nm, 100-mw NdYagfor Cy3; and a 633-nm, 35-mw HeNe for Cy5. Any two lasers may be turned on simulta-neously and their beams are delivered to the specimen via a single dichroic and an objectivelens (0.75 NA, 0.66-mm width). The emitted light, after passing back through the objectiveand primary dichroic, is focused through a confocal pinhole and through a secondary dich-roic onto two cooled photo multiplier tubes (PMTs), which operate in parallel for the two dif-ferent wavelengths. The stage is a standard computer-controled microscope stage capable of100 mm/s scans and 5 micron resolution. One or two standard 25

3

75 mm slides can bescanned at a time. At 100 mm/s, with 20 micron pixels, an area of 40

3

20 mm (capable ofprinting easily about 10 000 spots) can be scanned in about 20 min. About 10 pg/

m

L of eachspecies of cDNA can be reliably detected. Fig. 5 shows the images from two probes afterscanning and a combined color image.

Microarray image analysis programs, ArraySuite, as a set of extensions for IPLab Spec-trum for Macintosh computer (Scanalytics, Inc., Fairfax, VA, USA) and developed at NH-GRI, are used to analyze red and green hybridization intensities and red/green ratio for everygene on the array [9]. Features of ArraySuite include array alignment, DeArray, and Target-Locator. Fig. 6 illustrates the entire process of image analysis.

Fig. 4. Microarray reader in the Laboratory of Cancer Genetics, NHGRI, NIH. The reader is basically a computer-controled scanning confocal microscope with a triple laser illumination system. Scanning stage is shown on theleft insert. The optical system is shown on the right side.

C.C. Xiang, Y. Chen / Biotechnology Advances 18 (2000) 35–46

43

7. DNA microarray applications

7.1. Use of DNA microarrays for gene expression and discovery

Measuring transcript levels for thousands of genes in parallel is one of the more wide-spread applications of DNA microarray technology. Schena et al. [1] first described the high-capacity system of cDNA microarrays to monitor the expression of 45 Arabidopsis genes inparallel. Since then the cDNA microarray applications have been reported in many organ-isms, including plant [14], yeast [3,15], and human beings. The cDNA microarray technol-ogy was used to profile complex diseases and discover novel disease-related genes. DeRisi etal. [16] used the cDNA microarray technique to analyze gene expression patterns in humancancer. They demonstrated the tumorigenic properties of human melanoma cell line UACC-903 could be reversed by insertion of human chromosome 6. Heller et al. [17] studied geneexpression characteristic of the inflammatory disease rheumatoid arthritis and inflammatorybowel disease. They documented the stability of the cDNA microarray technology for profil-ing diseases and for identifying disease-related genes. The technology provided new targetsfor drug development and disease therapies, and in doing so allowed for improved treatmentof chronic diseases that were challenging because of their complexity. Welford et al. [18]used representational difference analysis coupled to cDNA microarray hybridization to de-

Fig. 5. Sample images. Image 1 is from Cy3-labeled probe. Image 2 is from Cy5-labeled probes. Images 3 is acombined color image representing differentially expressed genes. Arrows point the representative genes. R: red,G: green, Y: yellow.

44

C.C. Xiang, Y. Chen / Biotechnology Advances 18 (2000) 35–46

tect differentially expressed genes in primary tumor tissue. Their data showed that the use ofrepresentational difference analysis essentially provided an enriched library of differentiallyexpressed genes, while analysis of the library with microarray technology allowed rapid andreproducible screening of thousands of DNA molecules simultaneously. They further indi-cated that the coupling of two techniques in their system resulted in a large pool of differen-tially expressed genes.

7.2. Use of DNA microarrays for predicting various biochemical pathways

Schena et al. [19] monitored the expression of 1046 human cDNAs of unknown sequenceusing two-color differential expression analysis of heat shock and phorbol ester-regulatedgenes. They were the first to demonstrate the ability of cDNA microarrays to rapidly providedata for correlation of gene expression to biochemical pathways. The cDNA microarrayscontaining virtually all of the genes of

Saccharomyces cerevisiae

have been fabricated. De-Risi et al. [15] studied the metabolic and genetic control of gene expression on a genomicscale in yeast by using the yeast microarrays. They examined the effects of the diauxic shiftfrom anaerobic to aerobic metabolism under glucose limitation and the concomitant switchto ethanol as a carbon source. The significance of their study was that it mapped the changesin expression of genes with known function to their metabolic pathways and vividly showedwhich metabolic pathways were programmed by the shift. Spellman et al. [11] comprehen-sively identified cell cycle-regulated genes in yeast by cDNA microarray hybridization. Theyfound 800 genes that meet an objective minimum criterion for cell cycle regulation and halfof them respond to either G1 cyclin or B-type cyclin or both of these cyclins. Lyer et al. [20]

Fig. 6. Microarray data flow chart. The database starts with sample information, through image analysis, and endswith Internet access.

C.C. Xiang, Y. Chen / Biotechnology Advances 18 (2000) 35–46

45

studied temporal transcriptional program in the response of human fibroblasts to serum.They applied microarrays with 8600 different human genes to the study of growth controland cell cycle progression in humans. They clustered genes into groups on the basis of tem-poral patterns of expression of those genes. Their data indicated that many features of thetranscriptional program appeared to be related to the physiology of wound repair, which sug-gested that the fibroblasts play a larger and more important role in the complex multicellularresponse than had previously been appreciated. Galitski et al. [21] used microarray-basedgene expression analysis to identify genes showing ploidy-dependent expression in isogenic

Saccharomyces cerevisiae

strains that varied in ploidy from haploid to tetraploid. They foundthat those genes were induced or repressed in proportion to the number of chromosome sets,regardless of the mating type.

7.3. Use of DNA microarrays for drug discovery and development

The expression pattern of a gene provides indirect information about function. Knowledgeof highly selective gene expression, as well as sequence homology to a known gene family,could provide a convenient shortcut for implicating a target in a given pathway or disease.The cDNA microarray technology has helped many pharmaceutical companies to identifyappropriate targets for therapeutic intervention. DNA microarrays have also been used tomonitor changes in gene expression in response to drug treatment. Pietu et al. [22] identifiednovel gene transcripts preferentially expressed in human muscles by quantitative hybridiza-tion of a high density cDNA array. Among those genes, cathepsin K is a novel cysteine pro-tease that is expressed selectively in osteoclasts. The discovery led to the development ofdrugs to inhibit the cathepsin K. Marton [23] performed drug validation studies and identi-fied secondary drug target effects using DNA microarrays.

cDNA microarray or DNA chip technology is a powerful new approach for simulta-neously monitoring the relative expression of a large number of genes in a quantitative fash-ion. cDNA microarray technology is rapidly advancing. Its applications to gene discovery,gene expression, and mapping have been convincingly demonstrated. The suitability of thecDNA microarray for profiling diseases and for identifying disease-related genes has beenalso well documented. This novel technology could provide new targets for drug develop-ment and disease therapies. It may thus provide a useful link between gene sequences andclinical medicine for both human beings and animals. DNA microarrays seem likely to be-come a standard tool for both molecular biology research and clinical diagnostics.

References

[1] Schena M, Shalon D, Davis RW, Brown PO. Quantitative monitoring of gene expression patterns with acomplementary DNA microarray. Science 1995;270:467–70.

[2] Lockhart DJ. Expression monitoring by hybridization to high-density oligonucleotide arrays. Nature Bio-technol 1996;14:1675–80.

[3] Shalon D, Smith SJ, Brown PO. A DNA microarray system for analyzing complex DNA samples using two-color fluorescent probe hybridization. Genome Res 1996;6:639–45.

[4] Cheung VG, Morley M, Aguilar F, Massimi A, Kucherlapati R, Childs G. Making and reading microarrays.Nature Genetics 1999;21(Suppl.):15–9.

46

C.C. Xiang, Y. Chen / Biotechnology Advances 18 (2000) 35–46

[5] Velculescu VE, Zhang L, Vogelstein B, Kinzler KW. Serial analysis of gene expression. Science1995;270:484–7.

[6] Adams MD. Serial analysis of gene expression: ESTs get smaller. Bioassays 1996;18:261–2.[7] Ermolaeva O, Rastogi M, Pruitt KD, Schuler GD, Bittner ML, Chen Y, Simon R, Meltzer P, Trent JM, Bo-

guski MS. Data management and analysis for gene expression arrays. Nature Genetics 1998;20:19–23.[8] Bowtell DD. Options available—from start to finish—for obtaining expression data by microarray. Nat

Genet 1999;21(Suppl.):25–32.[9] Chen Y, Dougherty ER, Bittner ML. Ratio-based decisions and the quantitative analysis of cDNA microar-

ray images. J Biomedical Optics 1997;2:364–74.[10] Wen X, Fuhrman S, Michaels GS, Carr DB, Smith S, Barker JL, Somogyi JL. Large-scale temporal gene ex-

pression mapping of central nervous system development. Proc Natl Acad Sci 1998;95:334–9.[11] Spellman PT, Sherlock G, Zhang MQ, Iyer VR, Anders K, Eisen MB, Brown PO, Botstein D, Futcher B.

Comprehensive identification of cell cycle-regulated genes of the yeast

Saccharomyces cerevisiae

by mi-croarray hybridization. Molecular Biology of the Cell 1998;9:3273–97.

[12] Khan J, Simon R, Bittner M, Chen Y, Leighton SB, Pohida T, Smith PD, Jiang Y, Gooden GC, Trent JM,Meltzer PS. Gene expression profiling of alveolar rhabdomyosarcoma with cDNA microarrays. Cancer Res1998;58:5009–13.

[13] Eisen MB, Spellman PT, Brown PO, Botstein D. Cluster analysis and display of genome-wide expressionpatterns. Proc Natl Acad Sci USA 1998;95:14863–8.

[14] Schena M. Genome analysis with gene expression microarrays. Bioassays 1996;18:427–31.[15] DeRisi J, Vishwanath RL, Brown PO. Exploring the metabolic and genetic control of gene expression on a

genomic scale. Science 1997;278:680–6.[16] DeRisi J, Penland L, Brown PO, Bittner ML, Meltzer PS, Ray M, Chen Y, Su YA, Trent JM. Use of a cDNA

microarray to analyse gene expression patterns in human cancer. Nature Genetics 1996;14:457–60.[17] Heller RA, Schena M, Chai A, Shalon D, Bedilion T, Gilmore J, Woolley DE, Davis RW. Discovery and

analysis of inflammatory disease-related genes using cDNA microarray. Proc Natl Acad Sci USA1997;94:2150–5.

[18] Welford SM, Gregg J, Chen E, Garrison D, Sorensen PH, Denny CT, Nelson S. Detection of differentiallyexpressed genes in primary tumor tissues using representational differences analysis coupled to microarrayhybridization. Nucl Acids Res 1998;26:3059–65.

[19] Schena M, Shalon D, Heller R, Chai A, Brown PO, Davis RW. Parallel human genome analysis: Microar-ray-based expression monitoring of 1000 genes. Proc Natl Acad Sci USA 1996;93:10614–9.

[20] Lyer VR, Eisen MB, Ross DT, Schuler G, Moore T, Lee JCF, Trent JM, Staudt LM, Hudson J Jr, BoguskiMS, Lashkari D, Shalon D, Botstein D, Brown PO. The transcriptional program in the response of human fi-broblasts to serum. Science 1999;283:83–7.

[21] Galitski T, Saldanha AJ, Styles CA, Lander ES, Fink GR. Ploidy regulation of gene expression. Science1999;285:251–4.

[22] Pietu G, Alibert O, Guichard V, Lamy B, Bois F, Leroy E, Mariage-Samson R, Houlgatte R, Soularue P,Auffray C. Novel gene transcripts preferentially expressed in human muscles revealed by quantitative hy-bridization of a high-density cDNA array. Genome Res 1996;6:492–503.

[23] Marton MJ. Drug target validation and identification of secondary drug target effects using DNA microar-rays. Nature Med 1998;4:1293–301.