The completion of the sequencing of a large number of prokaryotic and eukaryotic genomes presents several challenges and opportunities, including the functional classification of predicted genes.
Microarray analysis promises to contribute to the functional annotation of genomes and has already provided a wealth of genome-wide expression data.
Much attention has been focused on experimental protocols for microarray studies, but the strategies for data analysis have a profound (and perhaps underappreciated) effect on the interpretation of the results.
Expression data from each experiment must first be normalized to account for systematic experimental variation, including unequal dye incorporation and detection efficiencies.
For comparison between experiments, data is often first filtered to select a subset or to exclude genes for which there is much missing data. A distance metric must then be chosen, which determines how we measure similarity between gene-expression patterns. Genes and experiments can then be grouped using various computational methods. Each step can influence how the expression data are grouped.
Clustering algorithms, which are the most widely used approaches to analysing gene expression, can be classified as hierarchical or non-hierarchical (self-organizing maps (SOMs), k-means clustering and principal component analysis), agglomerative (hierarchical) or divisive (k-means, SOMs), and supervised (support vector machine) or non-supervised (hierarchical and k-means clustering, SOMs).
A synthetic data set with well-defined relationships between genes is used to show the differences between some of these methods.
The choice of data analysis strategy should be influenced by the purpose of the microarray experiment, and the user's knowledge of the biology of the system under investigation.
Microarray experiments are providing unprecedented quantities of genome-wide data on gene-expression patterns. Although this technique has been enthusiastically developed and applied in many biological contexts, the management and analysis of the millions of data points that result from these experiments has received less attention. Sophisticated computational tools are available, but the methods that are used to analyse the data can have a profound influence on the interpretation of the results. A basic understanding of these computational tools is therefore required for optimal experimental design and meaningful data analysis.
This is a preview of subscription content, access via your institution
Subscribe to this journal
Receive 12 print issues and online access
$189.00 per year
only $15.75 per issue
Rent or buy this article
Prices vary by article type
Prices may be subject to local taxes which are calculated during checkout
Velculescu, V. E., Zhang, L., Vogelstein, B. & Kinzler, K. W. Serial analysis of gene expression. Science 270, 484–487 (1995).
Lockhart, D. J. et al. Expression monitoring by hybridization to high-density oligonucleotide arrays. Nature Biotechnol. 14, 1675–1680 (1996).
Schena, M., Shalon, D., Davis, R. W. & Brown, P. O. Quantitative monitoring of gene expression patterns with complementary DNA microarray. Science 270, 467–470 (1995).
Schena, M. et al. Parallel human genome analysis: microarray-based expression monitoring of 1000 genes. Proc. Natl Acad. Sci. USA 93, 10614–10619 (1996).
Wen, X. et al. Large-scale temporal gene expression mapping of central nervous system development. Proc. Natl Acad. Sci. USA 95, 334–339 (1998).This is one of the first analyses of large-scale gene expression — in this case, RT–PCR data — using clustering and data-mining techniques. It elegantly shows how integrating the results derived using various distance metrics can reveal different but meaningful patterns in the data.
Michaels, G. S. et al. Cluster analysis and data visualization of large-scale gene expression data. Pacific Symp. Biocomput. 1998, 42–53 (1998).
Eisen, M. B., Spellman, P. T., Brown, P. O. & Botstein, D. Cluster analysis and display of genome-wide expression patterns. Proc. Natl Acad. Sci. USA 95, 14863–14868 (1998).This is an excellent demonstration of the power of hierarchical clustering to the analysis of microarray data. The authors also provide software — Cluster and Treeview — which became the standard for analysing microarray data.
Weinstein, J. N. et al. An information-intensive approach to the molecular pharmacology of cancer. Science 275, 343–349 (1997).Weinstein and colleagues present one of the first and most elegant applications of hierarchical clustering and other data-mining and visualization techniques to the analysis of large-scale data in molecular biology.
Sokal, R. R. & Michener, C. D. A statistical method for evaluating systematic relationships. Univ. Kans. Sci. Bull. 38, 1409–1438 (1958).
Shannon, C. C. & Weaver, W. The Mathematical Theory of Communication (Illinois Univ. Press, Illinois, 1963).
Kohonen, T. Self Organizing Maps (Springer, Berlin, 1995).
Tamayo, P. et al. Interpreting patterns of gene expression with self-organizing maps: methods and application to hematopoietic differentiation. Proc. Natl Acad. Sci. USA 96, 2907–2912 (1999).Tamayo and colleagues use self-organizing maps (SOMs) to explore patterns of gene expression generated using Affymetrix arrays, and provide the GENECLUSTER implementation of SOMs.
Eisen, M. B. & Brown, P. O. DNA arrays for analysis of gene expression. Meth. Enzymol. 303, 179–205 (1999).
Hegde, P. et al. A concise guide to microarray analysis. Biotechniques 29, 548–560 (2000).
Boguski, M. S. & Schuler, G. D. ESTablishing a human transcript map. Nature Genet. 10, 369–371 (1995).
Quackenbush, J. et al. The TIGR gene indices: analysis of gene transcript sequences in highly sampled eukaryotic species. Nucleic Acids Res. 29, 159–164 (2001).
Burke, J., Wang, H., Hide, W. & Davison, D. B. Alternative gene form discovery and candidate gene selection from gene indexing projects. Genome Res. 8, 276–290 (1998).
Ermolaeva, O. et al. Data management and analysis for gene expression arrays. Nature Genet. 20, 19–23 (1998).
Sherlock, G. et al. The Stanford Microarray Database. Nucleic Acids Res. 29, 152–155 (2001).
Chen, Y., Dougherty, E. R. & Bittner, M. Ratio-based decisions and the quantitative analysis of cDNA microarray images. J. Biomed. Opt. 2, 364–374 (1997).
Heyer, L. J., Kruglyak, L. & Yooseph, S. Exploring expression data: identification and analysis of coexpressed genes. Genome Res. 9, 1106–1115 (1999).
Tavazoie, S., Hughes, J. D., Campbell, M. J., Cho, R. J. & Church, G. M. Systematic determination of genetic network architecture. Nature Genet. 22, 281–285 (1999).
Raychaudhuri, S., Stuart, J. M. & Altman, R. B. Principal components analysis to summarize microarray experiments: application to sporulation time series. Pac. Symp. Biocomput. 2000, 455–466 (2000).
Brown, M. P. et al. Knowledge-based analysis of microarray gene expression data by using support vector machines. Proc. Natl Acad. Sci. USA 97, 262–267 (2000).This paper shows the power of supervised techniques, in this case support vector machines, to provide additional insight into gene expression and function.
Golub, T. R. Molecular classification of cancer: class discovery and class prediction by gene expression monitoring. Science 286, 531–537 (1999).
Furey, T. S. et al. Support vector machine classification and validation of cancer tissue samples using microarray expression data. Bioinformatics 16, 906–914 (2000).
Hedenfalk, I. et al. Gene-expression profiles in hereditary breast cancer. N. Engl. J. Med. 344, 539–548 (2001).
Chatterjee, S. & Price, B. Regression Analysis by Example (John Wiley and Sons, New York, 1991).
Cleveland, W. S. & Devlin, S. J. Locally weighted regression: an approach to regression analysis by local fitting. J. Am. Stat. Assoc. 83, 596–610 (1988).
Sokal, R. R. & Sneath, P. H. A. Principles of Numerical Taxonomy (W. H. Freeman & Co., San Francisco, 1963).
Ward, J. H. Hierarchical grouping to optimize an objective function. J. Am. Stat. Assoc. 58, 236–244 (1963).
Cluster analysis was done using the The Institute for Genomic Research MeV software package developed by A. Sturn, A. I. Saeed and J.Q., which is available at http://pga.tigr.org/tools.shtml, along with the sample data set used here. The author also thanks A. Sturn, N. H. Lee, R. L. Malek and E. Snesrud for valuable discussions and comments. This work is supported by grants from the US National Science Foundation, the US National Cancer Institute, and the US National Heart, Lung, and Blood Institute.
PUBLIC EST SEQUENCES
DATA ANALYSIS TOOLS
META-LISTS OF OTHER AVAILABLE SOFTWARE
- CLUSTER ANALYSIS
The term 'cluster analysis' actually encompasses several different classification algorithms that can be used to develop taxonomies (typically as part of exploratory data analysis). Note that in this classification, the higher the level of aggregation, the less similar are members in the respective class.
The centroid of a cluster is the weighted average point in the multidimensional space; in a sense, it is the centre of gravity for the respective cluster.
A branching 'tree' diagram representing a hierarchy of categories on the basis of degree of similarity or number of shared characteristics, especially in biological taxonomy. The results of hierarchical clustering are presented as dendrograms, in which the distance along the tree from one element to the next represents their relative degree of similarity.
- NEURAL NETWORKS
Neural networks are analytic techniques modelled after the (proposed) processes of learning in cognitive systems and the neurological functions of the brain. Neural networks use a data 'training set' to build rules capable of making predictions or classifications on data sets.
- FACTOR ANALYSIS
Factor analysis is a data reduction and exploratory method similar to pincipal component analysis. Factor analysis techniques seek to reduce the number of variables and to detect structure in the relationships between elements in an analysis.
- PRINCIPAL COORDINATE ANALYSIS
Like principal component analysis, principal coordinate analysis seeks to reduce the dimensionality of a spatial representation of a data set by creating new coordinate axes that are a combination of the originals, and projecting the data onto those new axes.
A hyperplane is an N-dimensional analogy of a line or plane, which divides an 'N + 1' dimensional space into two.
- KERNEL FUNCTION
In support vector machines, the kernel function is a generalization of the distance metric; it measures the distance between two expression vectors as the data are projected into higher-dimensional space.
About this article
Cite this article
Quackenbush, J. Computational analysis of microarray data . Nat Rev Genet 2, 418–427 (2001). https://doi.org/10.1038/35076576