Microarray analysis has become a widely used technique for the study of gene-expression patterns on a genomic scale [1, 2]. Oligonucleotide and cDNA arrays have been utilized to study mRNA  and protein levels , to decipher protein-DNA interactions , to analyze the DNA copy number , to detect methylated sequences , and to analyze gene phenotypes in living mammalian cells . Microarrays represent a very complex, multi step technique involving array fabrication, labeling, hybridization, and data analysis. Currently, most laboratories are using either one labeled sample (Affymetrix microarrays) or two labeled samples (cDNA microarrays) for hybridizations, but several applications have been established were three color microarrays are used [9, 10]. State-of-the-art microarrays can have from several hundred up to tens of thousands of elements annotated by dozens of parameters. Information on details of the bench work, typically kept in lab notebooks or scattered files, as well as information regarding spotting, reliable tracking of the spotted molecules, scanning, and image quantification settings, is important for the computational analysis and reproducibility of experiments. Every step generates a wealth of data spanning tens of megabytes and in each of them errors may occur or protocols might need optimization to improve results. Moreover, all these information must be archived according to accepted scientific standards, which allow scientists to share common information and to make valid comparisons among experiments. For this reason the Microarray Gene Expression Data Society (MGED)  is focusing on establishing standards for microarray data annotation and exchange, facilitating the creation of microarray databases and related software implementing these standards. MGED is heavily promoting the sharing of high quality, well annotated data within the life sciences community. Their initiatives – MIAME (Minimum Information About a Microarray Experiment) , MGED Ontology , and MAGE-ML (MicroArray Gene Expression Markup Language)  – maximize the value of microarray data by permitting greater opportunities for sharing information within scientific groups and thus for discovery. These will ultimately affect the description, analysis, and management of all high throughput biological data.
The 'list of genes' resulting from microarray analysis is not the end of a microarray experiment. The major challenge is to assign biological function and to generate new hypotheses. The simplest way to find genes of potential biological interest is to search the normalized data for the highly expressed ones. Additionally, identifying patterns of gene expression and grouping genes into expression classes can provide greater insight into their biological relevance. For this purpose several supervised or unsupervised clustering algorithms like support vector machines (SVM), hierarchical clustering, k-means, self organizing map (SOM), or principal component analysis (PCA) are in use. The annotation of genes or gene clusters can be achieved by mapping them to the Gene Ontology (GO)  in order to provide insights into relevant molecular functions, biological processes, and cellular components . Another way to identify genes of biological interest is to map the normalized data or gene expression clusters  to known metabolic pathways as provided e.g. by KEGG  or BioCarta .
Several academic as well as commercial systems are available that address at least some of the needs such as laboratory information management systems (LIMS) , microarray databases [21–24] and repositories, normalization, clustering, pathway or GO mapping tools or expression analysis platforms . However, freely available systems which integrate all the aspects mentioned above are rare and may lack important issues like usability, scalability, or standardized interfaces. Furthermore, for such integrated systems it is desireable to use a uniform and state-of-the-art software architecture in order to enhance setup, maintenance and further development.
We have therefore developed a Microarray Analysis and Retrieval System (MARS) using latest Java 2 Platform, Enterprise Edtition (J2EE) software technology. MARS provides modules mandatory for microarray databases:
a laboratory information management system (LIMS) to keep track of information that accrues during the microarray production and biomaterial manipulation
MAGE-ML export of data for depositing to public repositories e.g. ArrayExpress , GEO 
For these components already existing projects [21, 23, 26] have been evaluated. Their advantages as well as disadvantages have been taken into account for the design of MARS. Widely used concepts have been taken into consideration and accepted standard libraries like MAGE-STK have been used whenever possible. Additionally, we extented this solid foundation and added novel features which can be highlighted as distinct advantages of the MARS system.
a quality management application storing necessary quality control parameters indispensable for high-quality microarray data
Web services to connect several well established tools such as normalization, clustering and pathway annotation applications
applications for microarray normalization, gene expression clustering, and pathway exploration that are tightly integrated into the microarray analysis pipeline
a novel, comprehensive, and Web based user management system to administrate institutes, groups, users, and their corresponding access rights