Mining for Microarray Gold

The use of high-throughput gene expression analysis technologies, such as microarrays, yields reams of data. Unfortunately, data stemming from competing and alternate technologies is often stored in a variety of formats, making comprehensive analyses difficult. Kirkland, Wash.-based Rosetta Inpharmatics Inc. offers a product that now enables scientists to join all of their gene expression data in a single location, and to mine that data for nuggets of gold unimaginable when the experiments were originally conducted. The Rosetta Resolver System, version 2.0, "provides storage, retrieval, and analysis mechanisms for large volumes of gene expression data from a wide variety of technologies," according to Douglas Bassett, senior director of biosoftware products at Rosetta. Thus, it can be used to integrate data from nylon arrays, microarrays, and SAGE experiments, for example. It also provides "powerful statistical algorithms and error models for each technology ... which allow us to predict, based on the past behavior of a certain gene expression platform technology, ... the quality statistics associated with each measurement," says Bassett. This means that the system can determine how much weight to assign each data point in subsequent analyses. The advantage of this approach is that every experiment added to the database increases the value of the entire database. Thus, whereas it was once common to analyze only a subset of the limited number of datasets stored in a single lab PC, it is now worthwhile to integrate all gene expression data in a single location, so that every data point ever produced in the lab can be utilized in the context of each analysis, says Bassett. A typical problem that can be addressed with the Resolver system is the analysis of a pharmacological compound whose target is unknown. By careful comparison of gene expression profiles from cells or tissues treated with the compound, data regarding the target pathway can be obtained. Similarly, information about an unknown gene can be gathered by data mining, again by comparison of expression profiles. The efficacy of this approach was demonstrated in a recent publication, in which the functions of eight uncharacterized genes were predicted by analysis of their expression profiles.1 Experimental analysis of mutants of these genes confirmed their placement into the predicted pathways. The Resolver system is client-server based, using a Sun enterprise server and a Java-based client running on Microsoft® Windows™. Data is input into the Resolver database in the eXtensible Markup Language (XML)-based Gene Expression Markup Language (GEML) format. Rosetta and others developed GEML "to provide a standard method of exchanging gene expression data along with the associated gene and experiment annotation."2 The not-for-profit Object Management Group (OMG), which provides a forum for the development and maintenance of information technology standards, is currently considering GEML for adoption as the standard gene expression data exchange format. NetGenics and the European Bioinformatics Institute submitted competing proposals. These proposals are now being merged into a final proposal entitled MAGE that represents the best components of the three initial proposals. Once the OMG officially adopts a specific standard, Rosetta will migrate its software tools to that language, says Bassett. Scientists working with disparate DNA sequencing apparatuses and formats can use the de facto GenBank sequence format, which allows researchers to exchange DNA sequence data, according to Bassett. But, he adds, "There is no such mechanism for gene expression data, so we are starting from square one." The proposed GEML format is designed to describe gene expression data in a rich format that includes background annotation and methodology in addition to the experimental results. Resolver can currently convert data from array systems developed by Agilent Technologies, Affymetrix, Genomics, Rosetta, BioDiscovery, and Molecular Dynamics into GEML format. Other formats can be accommodated via the development of a GEML format translator, which can be written either by the user, or by Rosetta Inpharmatics Profession Services if desired. Rosetta offers a free software tool, GEML Conductor, which can understand the GEML format. Akin to Adobe Acrobat Reader, Conductor is a free data visualization tool that allows researchers to share their data with colleagues and broadcast that data on the Web. Rosetta targets the Resolver system at "leading research labs in the gene expression field," says Bassett. Academic labs at Harvard University and the University of Washington currently use the product, as do researchers in the pharmaceutical industry. One industrial client is Research Triangle Park, N.C.-based Paradigm Genetics, which began using Resolver in December 2000. According to Patrick Hurban, a research scientist and leader of the gene expression profiling group at Paradigm, the company considered two factors in adopting Resolver: scalability, and the quality of the analytical subsystem. Researchers at Paradigm are integrating gene expression, phenotypic, and metabolic data to characterize gene function in plants and microbes. Overall, says Hurban, Paradigm has been pleased with the results. "The folks at Rosetta have been very good about working with us to make sure that we're happy with the database, and that we're getting out of the database what we intend to get out of it," he says. --Jeffrey M. Perkel (jperkel@the-scientist.com) References 1.T.R. Hughes et al., "Functional discovery via a compendium of expression profiles," Cell, 102:109-26, July 7, 2000. 2.Rosetta Inpharmatics, "Gene Expression Markup Language: A common data format for gene expression data and annotation interchange," www.geml.org. For More Information Rosetta Inpharmatics Inc. (425) 820-8900 www.rii.com

Mining for Microarray Gold

Interested in reading more?

Become a Member of