Computational approaches to analysis of DNA microarray data.

OBJECTIVES: To review the current state of the art in computational methods for the analysis of DNA microarray data. METHODS: The review considers methods of microarray data collection, transformation and representation, comparisons and predictions of gene expression from the data, their mechanistic analysis, related systems biology, and the application of clustering techniques. RESULTS: Functional genomics approaches have greatly increased the rate at which data on biological systems is generated, leading to corresponding challenges in analyzing the data through advanced computational techniques. The paper compares and contrasts the application of computational clustering for discovery, comparison, and prediction of gene expression classes, together with their evaluation and relation to mechanistic analyses of biological systems. CONCLUSION: Methods for assaying gene expression levels by DNA microarray experiments produce considerably more data than other techniques, and require a wide variety of computational techniques for identifying patterns of expression that may be biologically significant. These will have to be verified and validated by comparison to results from other methods, integrated with other systems data, and provide the feedback for further experimentation for testing mechanistic or other biological hypotheses.

From 'omes to biology.

Technologies that have emerged from the genome project have dramatically increased our ability to generate data on the way in which organisms respond to their environments, how they execute their programmes of development and growth, and how these are altered in the development of disease states. However, our ability to analyse these large datasets has not kept pace with our ability to generate them and consequently new strategies must be developed to address the issues associated with their analysis. One approach that we have employed quite successfully is to look at data from microarrays (or proteomics or metabolomics experiments) not as independent datasets, but rather as elements of a much larger body of biological information across various scales that must be integrated with, and interpreted within, the context of such ancillary data. Here we outline the general approach and provide three examples from published studies of the way in which we have applied this strategy.

Grape-seed derived procyanidins interfere with adipogenesis of 3T3-L1 cells at the onset of differentiation.

OBJECTIVE: Our group's previous results on the effects of a grape seed procyanidin extract (GSPE) on adipose metabolism showed that peroxisome proliferator-activated receptor-gamma (PPARgamma) plays a central role in the lipolytic effects of GSPE on adipocytes. Since PPARgamma2 is a main regulator of the differentiation process of adipocytes, we investigated whether GSPE affects the adipogenesis of 3T3-L1 cells. DESIGN: We performed a time point screening by treating 3T3-L1 cells with GSPE during the differentiation process for 24 h. MEASUREMENTS: Differentiation markers and differential gene expression due to GSPE treatment (using the microarray technique). RESULTS: Twenty four hour-GSPE treatment at the onset of differentiation reduces adipose-specific markers and maintains the expression of preadipocyte marker preadipocyte factor-1 (Pref-1) significantly elevated. These effects were not found in other time points. Microarray analysis of gene expression after GSPE treatment at the early stage of differentiation showed a modified gene expression profile in which cell cycle and growth-related genes were downregulated by GSPE. CONCLUSION: These results suggest that GSPE affects adipogenesis, mainly at the induction of differentiation, and that procyanidins may have a new role in which they impede the formation of adipose cells.

BmiGI: a database of cDNAs expressed in Boophilus microplus, the tropical/southern cattle tick.

We used an expressed sequence tag approach to initiate a study of the genome of the southern cattle tick, Boophilus microplus. A normalized cDNA library was synthesized from pooled RNA purified from tick larvae which had been subjected to different treatments, including acaricide exposure, heat shock, cold shock, host odor, and infection with Babesia bovis. For the acaricide exposure experiments, we used several strains of ticks, which varied in their levels of susceptibility to pyrethroid, organophosphate and amitraz. We also included RNA purified from samples of eggs, nymphs and adult ticks and dissected tick organs. Plasmid DNA was prepared from 11,520 cDNA clones and both 5' and 3' sequencing performed on each clone. The sequence data was used to search public protein databases and a B. microplus gene index was constructed, consisting of 8270 unique sequences whose associated putative functional assignments, when available, can be viewed at the TIGR website (http://www.tigr.org/tdb/tgi). A number of novel sequences were identified which possessed significant sequence similarity to genes, which might be involved in resistance to acaricides.

The TIGR Gene Indices: clustering and assembling EST and known genes and integration with eukaryotic genomes.

Although the list of completed genome sequencing projects has expanded rapidly, sequencing and analysis of expressed sequence tags (ESTs) remain a primary tool for discovery of novel genes in many eukaryotes and a key element in genome annotation. The TIGR Gene Indices (http://www.tigr.org/tdb/tgi) are a collection of 77 species-specific databases that use a highly refined protocol to analyze gene and EST sequences in an attempt to identify and characterize expressed transcripts and to present them on the Web in a user-friendly, consistent fashion. A Gene Index database is constructed for each selected organism by first clustering, then assembling EST and annotated cDNA and gene sequences from GenBank. This process produces a set of unique, high-fidelity virtual transcripts, or tentative consensus (TC) sequences. The TC sequences can be used to provide putative genes with functional annotation, to link the transcripts to genetic and physical maps, to provide links to orthologous and paralogous genes, and as a resource for comparative and functional genomic analysis.

Enrichment of gene-coding sequences in maize by genome filtration.

Approximately 80% of the maize genome comprises highly repetitive sequences interspersed with single-copy, gene-rich sequences, and standard genome sequencing strategies are not readily adaptable to this type of genome. Methodologies that enrich for genic sequences might more rapidly generate useful results from complex genomes. Equivalent numbers of clones from maize selected by techniques called methylation filtering and High C0t selection were sequenced to generate approximately 200,000 reads (approximately 132 megabases), which were assembled into contigs. Combination of the two techniques resulted in a sixfold reduction in the effective genome size and a fourfold increase in the gene identification rate in comparison to a nonenriched library.

Sequence analysis of a rainbow trout cDNA library and creation of a gene index.

Expressed sequence tag (EST) projects have produced extremely valuable resources for identifying genes affecting phenotypes of interest. A large-scale EST sequencing project for rainbow trout was initiated to identify and functionally annotate as many unique transcripts as possible. Over 45,000 5' ESTs were obtained by sequencing clones from a single normalized library constructed using mRNA from six tissues. The production of this sequence data and creation of a rainbow trout Gene Index eliminating redundancy and providing annotation for these sequences will facilitate research in this species.

Analysis of the mouse transcriptome based on functional annotation of 60,770 full-length cDNAs.

Only a small proportion of the mouse genome is transcribed into mature messenger RNA transcripts. There is an international collaborative effort to identify all full-length mRNA transcripts from the mouse, and to ensure that each is represented in a physical collection of clones. Here we report the manual annotation of 60,770 full-length mouse complementary DNA sequences. These are clustered into 33,409 'transcriptional units', contributing 90.1% of a newly established mouse transcriptome database. Of these transcriptional units, 4,258 are new protein-coding and 11,665 are new non-coding messages, indicating that non-coding RNA is a major component of the transcriptome. 41% of all transcriptional units showed evidence of alternative splicing. In protein-coding transcripts, 79% of splice variations altered the protein product. Whole-transcriptome analyses resulted in the identification of 2,431 sense-antisense pairs. The present work, completely supported by physical clones, provides the most comprehensive survey of a mammalian transcriptome so far, and is a valuable resource for functional genomics.

Use of RNA and genomic DNA references for inferred comparisons in DNA microarray analyses.

In most microarray assays, labeled cDNA molecules derived from reference and query RNA samples are co-hybridized to probes arrayed on a glass surface. Gene expression profiles are then calculated for each gene based on the relative hybridization intensities measured between the two samples. The most commonly used reference samples are typically isolates from a single representative RNA source (RNA-0) or pooled mixtures of RNA derived from a plurality of sources (RNA-p). Genomic DNA offers an alternative reference nucleic acid with a number of potential advantages, including stability, reproducibility, and a potentially uniform representation of all genes, as each unique gene should have equal representation in a haploid genome. Using hydrogen peroxide-treated Arabidopsis thaliana plants as a model, we evaluated genomic DNA and RNA-p as reference samples and compared expression levels inferred through the reference relative to unexposed plants with expression levels measured directly using an RNA-0 reference. Our analysis demonstrates that while genomic DNA can serve as a reasonable reference source for microarray assays, a much greater correlation with direct measurements can be achieved using an RNA-based reference sample.

Monitoring of representational difference analysis subtraction procedures by global microarrays.

Various approaches to the study of differential gene expression are applied to compare cell lines and tissue samples in a wide range of biological contexts. The compromise between focusing on only the important genes in certain cellular processes and achieving a complete picture is critical for the selection of strategy. We demonstrate how global microarray technology can be used for the exploration of the differentially expressed genes extracted through representational difference analysis (RDA). The subtraction of ubiquitous gene fragments from the two samples was demonstrated using cDNA microarrays including more than 32 000 spotted, PCR-amplified human clones. Hybridizations indicated the expression of 9100 of the microarray elements in a macrophage/foam cell atherosclerosis model system, of which many were removed during the RDA process. The stepwise subtraction procedure was demonstrated to yield an efficient enrichment of gene fragments overrepresented in either sample (18% in the representations, 86% after the first subtraction, and 88% after the second subtraction), many of which were impossible to detect in the starting material. Interestingly, the method allowed for the observation of the differential expression of several members of the low-abundant nuclear receptor gene family. We also observed a certain background level in the difference products of nondifferentially expressed gene fragments, warranting a verification strategy for selected candidate genes. The differential expression of several genes was verified by real-time PCR.

Minimum information about a microarray experiment (MIAME)-toward standards for microarray data.

Microarray analysis has become a widely used tool for the generation of gene expression data on a genomic scale. Although many significant results have been derived from microarray studies, one limitation has been the lack of standards for presenting and exchanging such data. Here we present a proposal, the Minimum Information About a Microarray Experiment (MIAME), that describes the minimum information required to ensure that microarray data can be easily interpreted and that results derived from its analysis can be independently verified. The ultimate goal of this work is to establish a standard for recording and reporting microarray-based gene expression data, which will in turn facilitate the establishment of databases and public repositories and enable the development of data analysis tools. With respect to MIAME, we concentrate on defining the content and structure of the necessary information rather than the technical format for capturing it.

Characterization of a Vibrio vulnificus LysR homologue, HupR, which regulates expression of the haem uptake outer membrane protein, HupA.

In Vibrio vulnificus, the ability to acquire iron from the host has been shown to correlate with virulence. Here, we show that the DNA upstream of hupA (haem uptake receptor) in V. vulnificus encodes a protein in the inverse orientation to hupA (named hupR). HupR shares homology with the LysR family of positive transcriptional activators. A hupA-lacZ fusion contained on a plasmid was transformed into Fur(-), Fur(+)and HupR(-)strains of V. vulnificus. The beta-galactosidase assays and Northern blot analysis showed that transcription of hupA is negatively regulated by iron and the Fur repressor in V. vulnificus. Under low-iron conditions with added haemin, the expression of hupA in the hupR mutant was significantly lower than in the wild-type. This diminished response to haem was detected by both Northern blot and hupA-lacZ fusion analysis. The haem response of hupA in the hupR mutant was restored to wild-type levels when complemented with hupR in trans. These studies suggest that HupR may act as a positive regulator of hupA transcription under low-iron conditions in the presence of haemin.

Identification of tumor markers in models of human colorectal cancer using a 19,200-element complementary DNA microarray.

Metastasis represents a crucial transition in disease development and progression and has a profound impact on survival for a wide variety of cancers. Cell line models of metastasis have played an important role in developing our understanding of the metastatic process. We used a 19,200-element human cDNA microarray to profile transcription in three paired cell-line models of colorectal tumor metastasis. By correlating expression patterns across these cell lines, we have identified 176 genes that appear to be differentially expressed (greater than 2-fold) in all highly metastatic cell lines relative to their reference. An analysis of these genes reiterates much of our understanding of the metastatic process and suggests additional genes, many of previously uncharacterized function, that may be causatively involved in, or at least prognostic of, metastasis. Northern analysis of a limited number of these genes validates the observed pattern of expression and suggests that further investigation and functional characterization of the identified genes is warranted.

The power of public access: the human genome project and the scientific process.

The scientific process, and scientific progress, require a critical examination of all published reports. Recent publications detailing errors in the draft human genome sequence are an integral part of our quest to better understand nature and demonstrate the value of free access to scientific data.

Exploring the transcriptome of the malaria sporozoite stage.

Most studies of gene expression in Plasmodium have been concerned with asexual and/or sexual erythrocytic stages. Identification and cloning of genes expressed in the preerythrocytic stages lag far behind. We have constructed a high quality cDNA library of the Plasmodium sporozoite stage by using the rodent malaria parasite P. yoelii, an important model for malaria vaccine development. The technical obstacles associated with limited amounts of RNA material were overcome by PCR-amplifying the transcriptome before cloning. Contamination with mosquito RNA was negligible. Generation of 1,972 expressed sequence tags (EST) resulted in a total of 1,547 unique sequences, allowing insight into sporozoite gene expression. The circumsporozoite protein (CS) and the sporozoite surface protein 2 (SSP2) are well represented in the data set. A BLASTX search with all tags of the nonredundant protein database gave only 161 unique significant matches (P(N) < or = 10(-4)), whereas 1,386 of the unique sequences represented novel sporozoite-expressed genes. We identified ESTs for three proteins that may be involved in host cell invasion and documented their expression in sporozoites. These data should facilitate our understanding of the preerythrocytic Plasmodium life cycle stages and the development of preerythrocytic vaccines.

Effects of ischemia on gene expression.

Microarray gene expression technology has recently made it feasible to characterize the RNA expression of thousands of genes across numerous tissue samples. We hypothesized that the warm ischemia commonly associated with the surgical extirpation of human tissue would have significant effects on gene expression profiles. To quantitate the effects of warm ischemia on human tissue, we rapidly dissected normal mucosa from a human colon cancer specimen. The specimen was divided and maintained at room temperature until snap-frozen in liquid nitrogen. Aliquots of tissue were frozen at times 5, 10, 15, 20, 40, and 60 min after extirpation. Spotted microarrays composed of 2400 distinct elements were used to assay mRNA derived from each time point in triplicate. Eisen's hierarchical clustering methodology and Bayesean statistical methods were then used to assay the effects of warm ischemia on gene expression. Application of time-course statistical models suggest that three patterns were induced by ischemia, accounting for 68.2, 17.8, and 13.4% of the evaluable genes, respectively. Pattern I corresponds to an average change of 27% over 60 min from 5 min baseline level of expression and 63.8% of the genes with at least 80% probability of membership in this pattern show average increases in expression over 60 min. The remainder decrease on average. Pattern II genes show the least ischemia-related effects, demonstrating an average change of only 12% over 60 min. In contrast to pattern I, we find that 67.5% of the genes with at least 80% probability of membership in this pattern are decreasing in expression on average over time. The remaining 32.5% in this pattern increase an average of 12% over 60 min. Finally, pattern III genes (13.4% of the sample) show the greatest sensitivity to ischemia, changing an average of 50% over 60 min, with about the same number increasing as are decreasing. Fold changes in RNA over- or under-expression were observed up to greater than 20-fold. Warm ischemia associated with the surgical extirpation of human tissues has significant effects on gene expression. These data support the careful monitoring of ischemic time for tissues harvested for the purpose of gene profiling.

Computational analysis of microarray data.

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.

Microarray analysis of the in vivo effects of hypophysectomy and growth hormone treatment on gene expression in the rat.

Complementary DNA microarrays containing 3000 different rat genes were used to study the consequences of severe hormonal deficiency (hypophysectomy) on the gene expression patterns in heart, liver, and kidney. Hybridization signals were seen from a majority of the arrayed complementary DNAs; nonetheless, tissue-specific expression patterns could be delineated. Hypophysectomy affected the expression of genes involved in a variety of cellular functions. Between 16-29% of the detected transcripts from each tissue changed expression level as a reaction to this condition. Chronic treatment of hypophysectomized animals with human GH also caused significant changes in gene expression patterns. The study confirms previous knowledge concerning certain gene expression changes in the above-mentioned situations and provides new information regarding hypophysectomy and chronic human GH effects in the rat. Furthermore, we have identified several new genes that respond to GH treatment. Our results represent a first step toward a more global understanding of gene expression changes in states of hormonal deficiency.