A relational schema for both array-based and SAGE gene expression experiments

Navigating public microarray databases

With the ever-escalating amount of data being produced by genome-wide microarray studies, it is of increasing importance that these data are captured in public databases so that researchers can use this information to complement and enhance their own studies. Many groups have set up databases of expression data, ranging from large repositories, which are designed to comprehensively capture all published data, through to more specialized databases. The public repositories, such as ArrayExpress at the European Bioinformatics Institute contain complete datasets in raw format in addition to processed data, whilst the specialist databases tend to provide downstream analysis of normalized data from more focused studies and data sources. Here we provide a guide to the use of these public microarray resources.

The MGED ontology: a framework for describing functional genomics experiments

The Microarray Gene Expression Data (MGED) society was formed with an initial focus on experiments involving microarray technology. Despite the diversity of applications, there are common concepts used and a common need to capture experimental information in a standardized manner. In building the MGED ontology, it was recognized that it would be impractical to cover all the different types of experiments on all the different types of organisms by listing and defining all the types of organisms and their properties. Our solution was to create a framework for describing microarray experiments with an initial focus on the biological sample and its manipulation. For concepts that are common for many species, we could provide a manageable listing of controlled terms. For concepts that are species-specific or whose values cannot be readily listed, we created an ‘OntologyEntry’ concept that referenced an external resource. The MGED ontology is a work in progress that needs additional instances and particularly needs constraints to be added. The ontology currently covers the experimental sample and design, and we have begun capturing aspects of the microarrays themselves as well. The primary application of the ontology will be to develop forms for entering information into databases, and consequently allowing queries, taking advantage of the structure provided by the ontology. The application of an ontology of experimental conditions extends beyond microarray experiments and, as the scope of MGED includes other aspects of functional genomics, so too will the MGED ontology.

Microarray analysis of the developing cortex

Abnormal development of the prefrontal cortex (PFC) is associated with a number of neuropsychiatric disorders that have an onset in childhood or adolescence. Although the basic laminar structure of the PFC is established in utero, extensive remodeling continues into adolescence. To map the overall pattern of changes in cortical gene transcripts during postnatal development, we made serial measurements of mRNA levels in mouse PFC using oligonucleotide microarrays. We observed changes in mRNA transcripts consistent with known postnatal morphological and biochemical events. Overall, most transcripts that changed significantly showed a progressive decrease in abundance after birth, with the majority of change between postnatal weeks 2 and 4. Genes with cell proliferative, cytoskeletal, extracellular matrix, plasma membrane lipid/transport, protein folding, and regulatory functions had decreases in mRNA levels. Quantitative PCR verified the microarray results for six selected genes: DNA methyltransferase 3A (Dnmt3a), procollagen, type III, alpha 1 (Col3a1), solute carrier family 16 (monocarboxylic acid transporters), member 1 (Slc16a1), MARCKS-like 1 (Marcksl1), nidogen 1 (Nid1) and 3-hydroxybutyrate dehydrogenase (heart, mitochondrial) (Bdh).

Evaluation of various housekeeping genes for their applicability for normalization of mRNA expression in dioxin-treated rats

Quantitative reverse transcription-polymerase chain reaction (qRT-PCR) is an extremely sensitive, convenient and rapid method to measure mRNA levels in cells and tissues, and is gaining popularity in toxicology. To correct for sample-to-sample variation, normalization of the expression data is required. The conventional way to perform normalization is to select a reference gene whose expression is believed to remain stable across all experimental conditions, then relate the concentrations of gene(s) of interest to those of this housekeeping gene. Since recent evidence shows that some housekeeping genes are actually not as refractory to experimental manipulations as previously thought, we validated a large number (18) of commonly used housekeeping genes for acute toxicity studies of 2,3,7,8-tetrachlorodibenzo-p-dioxin (TCDD), an extremely potent environmental toxin known to regulate a wide variety of genes. Microarray and qRT-PCR analyses coherently demonstrated that about 50% of the housekeeping genes examined were responsive to TCDD in rat liver with the magnitudes of change up to nearly 10-fold. Extension of the study to spleen and hypothalamus verified that phosphoglycerate kinase 1 (Pgk1) and glyceraldehyde-3-phosphate dehydrogenase (GAPDH) retained their basal expression levels in all experimental settings, although body weight loss-generated repression may mask a slight induction of GAPDH by TCDD in liver. These findings show that normalization genes for qRT-PCR must be carefully validated in advance, especially if the study involves a potent modifier of gene expression.

Aryl hydrocarbon receptor regulates distinct dioxin-dependent and dioxin-independent gene batteries

Conventional biochemical and molecular techniques identified previously several genes whose expression is regulated by the aryl hydrocarbon receptor (AHR). We sought to map the complete spectrum of AHR-dependent genes in male adult liver using expression arrays to contrast mRNA profiles in Ahr-null mice (Ahr(-/-)) with those in mice with wild-type AHR (Ahr(+)(/)(+)). Transcript profiles were determined both in untreated mice and in mice treated 19 h earlier with 1000 microg/kg 2,3,7,8-tetrachlorodibenzo-p-dioxin (TCDD). Expression of 456 ProbeSets was significantly altered by TCDD in an AHR-dependent manner, including members of the classic AHRE-I gene battery, such as Cyp1a1, Cyp1a2, Cyp1b1, and Nqo1. In the absence of exogenous ligand, AHR status alone affected expression of 392 ProbeSets, suggesting that the AHR has multiple functions in normal physiology. In Ahr(-/-) mice, only 32 ProbeSets exhibited responses to TCDD, indicating that the AHR is required for virtually all transcriptional responses to dioxin exposure in liver. The flavin-containing monooxygenases, Fmo2 and Fmo3, considered previously to be uninducible, were highly induced by TCDD in an AHR-dependent manner. The estrogen receptor alpha as well as two estrogen-receptor-related genes (alpha and gamma) exhibit AHR-dependent expression, thereby extending cross-talk opportunities between the intensively studied AHR and estrogen receptor pathways. p53 binding sites are over-represented in genes down-regulated by TCDD, suggesting that TCDD inhibits p53 transcriptional activity. Overall, our study identifies a wide range of genes that depend on the AHR, either for constitutive expression or for response to TCDD.

Development of a microarray assay that measures hybridization stoichiometry in moles

Microarray data is most useful when it can be compared with other genetic detection technologies. In this report, we designed a microarray assay format that transforms raw data into a defined scientific unit (i.e., moles) by measuring the amount of array feature present and the cDNA sequence hybridized. This study profiles a mouse reference universal RNA sample on a microarray consisting of PCR products. In measuring array features, a labeled DNA sequence was designed that hybridizes to a conserved sequence that is present in every array feature. To measure the amount of cDNA sample hybridized, the RNA sample was processed to ensure consistent dye to DNA ratio for every labeled target cDNA molecule, using labeled branched dendrimers rather than by incorporation. A dye printing assay was then performed in order to correlate molecules of cyanine dye to signal intensity. We demonstrate that by using this microarray assay design, raw data can be transformed into defined scientific units, which will facilitate interpretation of other experiments, such as data deposited at the Gene Expression Omnibus and ArrayExpress.

Maintaining data integrity in microarray data management

Gene expression microarrays are a relatively new technology, dating back just a few years, yet they have already become a very widely used tool in biology, and have evolved to a wide range of applications well beyond their original design intent. However, while the use of microarrays has expanded, and the issues of performance optimization have been intensively studied, the fundamental issue of data integrity management has largely been ignored. Now that performance has improved so greatly, the shortcomings of data integrity control methods constitute a greater percent of the stumbling blocks for investigators. Microarray data are cumbersome, and the rule up to this point has mostly been one of hands-on transformations, leading to human errors which often have dramatic consequences. We show in this review that the time lost on such mistakes is enormous and dramatically affects results; therefore, mistakes should be mitigated in any way possible. We outline the scope of the data integrity issue, to survey some of the most common and dangerous data transformations, and their shortcomings. To illustrate, we review some case studies. We then look at the work done by the research community on this issue (which admittedly is meager up to this point). Some data integrity issues are always going to be difficult, while others will become easier-one of our goals is to expedite the use of integrity control methods. Finally, we present some preliminary guidelines and some specific approaches that we believe should be the focus of future research.

Absolute expression values for mouse transcripts: re-annotation of the READ expression database by the use of CAGE and EST sequence tags

The RIKEN expression array database (READ) provides comprehensive gene expression data for the mouse, which were obtained as relative values from microarray double-staining experiments with E17.5 mRNA as common reference. To assign absolute expression values for mouse transcripts within READ, we applied the E17.5 reference sample to CAGE (cap analysis of gene expression) and expressed sequence tag (EST) high-throughput tag sequencing. Newly assigned values within the READ database were validated by comparison to expression data from serial analysis of gene expression, CAGE and EST experiments. These experiments confirmed the great significance of the absolute expression values within the improved READ database. The new Absolute READ database on absolute expression data is available under.

Fidelity and enhanced sensitivity of differential transcription profiles following linear amplification of nanogram amounts of endothelial mRNA

Although mRNA amplification is necessary for microarray analyses from limited amounts of cells and tissues, the accuracy of transcription profiles following amplification has not been well characterized. We tested the fidelity of differential gene expression following linear amplification by T7-mediated transcription in a well-established in vitro model of cytokine [tumor necrosis factor alpha (TNFalpha)]-stimulated human endothelial cells using filter arrays of 13,824 human cDNAs. Transcriptional profiles generated from amplified antisense RNA (aRNA) (from 100 ng total RNA, approximately 1 ng mRNA) were compared with profiles generated from unamplified RNA originating from the same homogeneous pool. Amplification accurately identified TNFalpha-induced differential expression in 94% of the genes detected using unamplified samples. Furthermore, an additional 1,150 genes were identified as putatively differentially expressed using amplified RNA which remained undetected using unamplified RNA. Of genes sampled from this set, 67% were validated by quantitative real-time PCR as truly differentially expressed. Thus, in addition to demonstrating fidelity in gene expression relative to unamplified samples, linear amplification results in improved sensitivity of detection and enhances the discovery potential of high-throughput screening by microarrays.

ToxoDB: accessing the Toxoplasma gondii genome

ToxoDB (http://ToxoDB.org) provides a genome resource for the protozoan parasite Toxoplasma gondii. Several sequencing projects devoted to T. gondii have been completed or are in progress: an EST project (http://genome.wustl.edu/est/index.php?toxoplasma=1), a BAC clone end-sequencing project (http://www.sanger.ac.uk/Projects/T_gondii/) and an 8X random shotgun genomic sequencing project (http://www.tigr.org/tdb/e2k1/tga1/). ToxoDB was designed to provide a central point of access for all available T. gondii data, and a variety of data mining tools useful for the analysis of unfinished, un-annotated draft sequence during the early phases of the genome project. In later stages, as more and different types of data become available (microarray, proteomic, SNP, QTL, etc.) the database will provide an integrated data analysis platform facilitating user-defined queries across the different data types.

Microarray databases: standards and ontologies

A single microarray can provide information on the expression of tens of thousands of genes. The amount of information generated by a microarray-based experiment is sufficiently large that no single study can be expected to mine each nugget of scientific information. As a consequence, the scale and complexity of microarray experiments require that computer software programs do much of the data processing, storage, visualization, analysis and transfer. The adoption of common standards and ontologies for the management and sharing of microarray data is essential and will provide immediate benefit to the research community.

Gene expression profiling with DNA microarrays: advancing our understanding of psychiatric disorders

DNA microarray transcriptome profiling of the postmortem brain opens novel horizons in understanding molecular changes associated with complex psychiatric disorders. With careful analysis and interpretation of microarray data we are uncovering previously unknown, expression patterns that maybe subject-specific and pivotal in understanding the disease process. In our recent studies, analyses of the prefrontal cortex of subjects with schizophrenia and matched controls uncovered complex changes in the expression of genes related to presynaptic secretory release, GABAergic and glutamatergic transmission, metabolic pathways, myelination, as well as cAMP and phosphoinositol second messenger systems. Our goal will be to integrate this expression data within the context of the relevant anatomical, biochemical, molecular, imaging and clinical findings.

Comparison of different labeling methods for two-channel high-density microarray experiments

In this report we evaluate three methods for labeling nucleic acids to be hybridized to a cDNA microarray: direct labeling, indirect amino-allyl labeling, and the dendrimer labeling method (Genisphere). The dendrimer method requires the smallest quantity of sample, 2.5 microg of total RNA compared with 20 microg with the direct or indirect methods. Therefore, we wanted to know whether the performance of the dendrimer method is comparable to the other methods, or whether significant information is lost. Performance can be considered in terms of sensitivity, dynamic range, and reproducibility of the quantitative signals for gene intensity. We compared the three labeling methods by generating three sets of eight self-to-self hybridizations using the same total RNA sample in all cases ("replicate study"). In our analysis, we controlled for the effects of print-tip and background subtraction biases. We also performed a smaller study, namely, a dilution series study with five dilution points per labeling method, to evaluate one aspect of predictive ability. From the replicate study, the dendrimer method appeared to perform as well, and often better, with respect to reproducibility and ability to detect expression. However, in the dilution series study, this method was outperformed by the other two in terms of predictive ability and did not perform very well. These findings are helping to guide our decisions on what labeling method to use for subsequent studies, based on the purpose of a specific study and its limitations in terms of available material.

PlasmoDB: the Plasmodium genome resource. An integrated database providing tools for accessing, analyzing and mapping expression and sequence data (both finished and unfinished)

PlasmoDB (http://PlasmoDB.org) is the official database of the Plasmodium falciparum genome sequencing consortium. This resource incorporates finished and draft genome sequence data and annotation emerging from Plasmodium sequencing projects. PlasmoDB currently houses information from five parasite species and provides tools for cross-species comparisons. Sequence information is also integrated with other genomic-scale data emerging from the Plasmodium research community, including gene expression analysis from EST, SAGE and microarray projects. The relational schemas used to build PlasmoDB [Genomics Unified Schema (GUS) and RNA Abundance Database (RAD)] employ a highly structured format to accommodate the diverse data types generated by sequence and expression projects. A variety of tools allow researchers to formulate complex, biologically based queries of the database. A version of the database is also available on CD-ROM (Plasmodium GenePlot), facilitating access to the data in situations where Internet access is difficult (e.g. by malaria researchers working in the field). The goal of PlasmoDB is to enhance utilization of the vast quantities of data emerging from genome-scale projects by the global malaria research community.

Functional genomics: high-throughput mRNA, protein, and metabolite analyses

A tremendous amount of DNA sequence information is now available to scientists and engineers. These DNA sequences provide the foundation for studying how the genome of an organism is functioning and they are particularly useful for metabolic engineers interested in manipulating plants for the production of chemicals and enzymes. Functional genomics relies on high-throughput techniques for measuring the mRNA (the transcriptome), protein (the proteome), and metabolite (the metabolome) components of plants as well as their organs and tissues. Microarray technologies, recent advances in protein mass spectrometry, and high-throughput metabolite analyses are beginning to provide detailed information on the total mRNA, protein, and metabolite components of plants. This knowledge will allow scientists to monitor changes in proteins and metabolites in plants. Ultimately, it may allow them to discover new metabolic pathways and to model metabolic and regulatory networks in plants.

Sharing gene expression data: an array of options

Sharing of microarray data has many advantages for the scientific and biomedical community, and should be advocated by neuroscience journals. The goals of sharing are manifold, and include improving analysis and confidence in results, and facilitating global comparisons between experiments, while at the same time, not penalizing those who share. The sharing of microarray data poses unique challenges relative to more generic data such as DNA sequences. These challenges are surmountable, and various sharing formats are possible. Centralized non-commercial databases are being developed to facilitate this process.