Completion and annotation of the Arabidopsis thaliana genome represented a major step in plant genetic research 1. This knowledge enabled gene prediction, assignment of functional categories and gave an opportunity to study gene and chromosome organization including the distribution of transposable elements. Finally it has enabled the characterization of global gene expression patterns at the transcriptome level at different developmental stages and under various physiological and stress conditions. Efforts to reveal the biological functions of thousand of genes and their integration into proteome, metabolome and interactome networks has become the principal focus of many studies and represents key objective of the 2010 Project 2

A number of efficient and accurate gene expression analysis technologies to determine the expression levels of individual genes have been widely exploited in recent decades (Northern blot analysis, quantitative reverse transcription-PCR, cDNA library screening). Most of these methods enable analysis of the expression of single or relatively few selected genes. For the discovery of partial or whole gene functional or regulatory networks, the development of high-throughput technologies is essential with genome-wide transcriptomic studies providing a major input 3. Several such methods have been developed including, cDNA fingerprinting 4, serial analysis of gene expression – SAGE 5, massively parallel signature sequencing – MPSS 6, high-density DNA oligonucleotide probe microarrays 78 or cDNA arrays 9. DNA microarray technologies are among the most frequently used methods for parallel global analysis of gene expression. These methods are based on the principle of selective and differential hybridization between sample target molecules and immobilized DNA probes. Hybridisation to probes arrayed on a solid surface report the relative abundance of DNA or RNA target molecules by fluorescent signal detection 1011. Microarray technologies now belong to the standard functional genomics toolbox 1213 and have undergone massive development leading to increased genome coverage, accuracy and reliability. Whole Genome microarrays developed by Affymetrix (Santa Clara, CA, USA) in collaboration with Syngenta represented the first standard in genome wide transcriptomic studies in plants. Whole genome Affymetrix ATH1 GeneChips cover about 76% of the Arabidopsis thaliana genes 14. Moreover, the introduction of the Minimum Information About Microarray experiments (MIAME) as standard documentation for array experiments and in transcriptomic databases, increasing the value and comparability of microarray data 15.

The number of experiments exploiting microarray technology has markedly increased in recent years. Not surprisingly, there are potential difficulties in navigating between different available data sets. Microarray expression data are deposited on servers, many of which are publicly accessible. Public plant microarray data are deposited in several databases including ArrayExpress 16, GEO 17, NASCarrays 18 and the Stanford Microarray Database 192021. Currently these databases store several thousands of individual datasets and some of these offer on-line tools for data normalization, filtering, statistical testing and pattern discovery 2223242526.

In parallel with the rapid accumulation of transcriptomic data, on-line analysis tools are being introduced to simplify their use. Global statistical data analysis methods contribute to the identification of overall gene expression patterns and to query and compose new working hypotheses based on these findings 11122728. More recently, these applications are being supplemented with more specialized products offering visualization and specific data mining tools. Genevestigator, Botany Array Resource, Arabidopsis co-expression tool, and Expression Profiler offer Web-based tools to analyse large microarray datasets. Genevestigator offers two types of queries: a gene-centric approach and a genome-centric approach, which are represented by several analysis tools; Gene Correlator, Gene Atlas, Gene Chronologer, Response Viewer and the Meta-Analyzer, that is among the most sophisticated complex amongst available microarray analysis toolboxes 2930. Botany Array Resource offers similar services supplemented with tools for discovery and analyses of cis-elements in promoters 31. Expression Profiler (EP) provides tools for hierarchical and K-groups clustering, clustering comparison, similarity search or the signature algorithm 3233. The more specialized PathoPlant database on plant-pathogen interactions and components of signal transduction pathways related to plant pathogenesis also harbors gene expression data from Arabidopsis thaliana microarray experiments to enable searching for specific genes regulated upon pathogen infection or elicitor treatment 3435. Finally, Arabidopsis Co-Expression Tool (ACT) allows users to identify genes with expression patterns correlated across selected experiments or the complete data set and offers the novel clique finder tool 363738.

In this article we introduce Arabidopsis Gene Family Profiler (arabidopsisGFP, aGFP), a web-based gene expression database with visualization tools. During programming, we took into account that for many microarray data users, extraction of global expression patterns of single genes or gene families can be time-consuming and its visualization difficult. Moreover, the use of various normalization algorithms in individual experiments makes direct comparison of genes of interest within various datasets uncertain. To solve these issues, we developed aGFP to provide the user with two normalization and gene detection algorithms and a "virtual plant" Arabidopsis Gene Family Profiler facility that enables users to obtain a global expression profile of user specified and/or pre-defined gene families. These attributes of aGFP contribute a useful resource for the rapid bioinformatic analysis of Arabidopsis gene expression data through comparative expression profile analysis in a gene family-based context.

The Arabidopsis Gene Family Profiler (aGFP) database was designed to give users the possibility to visualize expression patterns of individual genes, pre-defined gene families or user-defined gene sets in various tissues and at different developmental stages of wild type Arabidopsis thaliana plants. aGFP largely exploits microarray experiments obtained through the NASC AffyWatch transcriptomics service 39. We adopted the general concept "from simple to complex". In the first approximation, an arithmetical mean expression signal from multiple experiments is displayed. In subsequent steps the user can choose to display expression data for individual plant organs or tissues at particular growth stages. This is accompanied by the option of progressive replacement of arithmetical means by individual expression values. So the user has the option to choose the different levels of visualization to suit needs. Finally, the user can switch from "virtual plant" visualization to a simple bar chart (standard or log-scaled) or tabulated display and can browse through individual experiments down to normalized or even raw data extracted from individual gene chips. Gene family data can also be visualized as a colorized spot chart.

Although the idea of web-based database tools is not novel, aGFP database offers a quick and interactive display of gene expression profiles using the virtual plant facility as well as alternate more conventional outputs. A novel feature of aGFP is that it enables the evaluation of the impact of normalization procedures on microarray expression data as well the possibility of rapid definition of user-defined families or gene groups. Simultaneously, aGFP serves as a facile and synoptic developmental reference guide for expression profiles of individual genes or gene families in wild-type Arabidopsis thaliana plants.

arabidopsisGFP is a microarray expression database of wild type Arabidopsis thaliana plants grown under physiological conditions. It gathers data from experiments using Affymetrix ATH1 whole genome arrays with two or more biological replicates. From the outset, it has been created as intuitive user-oriented web-tool employing a "general-to-specific" concept enabling the user to obtain certain amount of information at every step with progressive specification and refinement. The aGFP database contains several gene selection and grouping tools including pre-defined gene families. It also provides the user with different gene expression visualization options including a unique "virtual plant" graphic display. Easy switching of visualization options gives the user the possibility to rapidly select the most suitable form of data presentation. A novel advantage of the aGFP database is the provision of alternative normalization treatments of microarray data using statistical (MAS5.0) and empirical algorithms (MAS4.0). Together with the facile switch between these detection algorithms it provides the opportunity to instantly assess the reliability of gene expression data. Arabidopsis Gene Family Profiler represents a versatile tool for facile visualization of transcriptomic data that can be exploited in genome-led queries of gene and gene family functions and regulation.

The aGFP database is freely accessible and its concept offers the possibility to extract and visualize expression profiles of individual genes, gene sets, gene families or gene superfamilies from a broad spectrum of microarray experiments covering various Arabidopsis organs, tissues and developmental stages. For these purposes, an innovative graphic concept of the "virtual plant" was introduced representing a clear and simple visualization of gene expression profiles in a morphological and developmental context. The arabidopsisGFP database is accessible at http://agfp.ueb.cas.cz/.

ND, DT and DH defined the concept of arabidopsisGFP database. ND and PH programmed all scripts. DH is responsible for data normalisation. DR worked on gene families and superfamilies definition. DH, DT, ND and BH specified the data visualization concepts and BH drew the "virtual plant". All authors read and approved the final manuscript.