The retina is a thin and highly structured layer of neuronal cells that lines the back of eye. Its main function is to convert light energy into an interpretable signal for cortical cells in the brain. The retina has two components – an inner neurosensory retina and an outer retinal pigment epithelium (RPE), which together form the structural and functional basis for visual perception.

The retina consists of several cell types, of which neural cells predominate. Photoreceptors, bipolar and ganglion cells are three principal neuron cell types whose activity is modulated by other groups of cells, such as horizontal and amacrine cells 1. Defects in any of the above-mentioned cell types can lead to a variety of retinal diseases, including age-related macular degeneration (AMD), retinitis pigmentosa (RP), Leber congenital amaurosis (LCA) and glaucoma. These diseases may cause partial visual loss or complete blindness, depending on the severity.

The recent progress in genomic approaches has now led to an increase in the number of transgenic and knockout animal models that can be used to investigate the role of specific genes in retinal function and related disorders in humans, e.g., rd1 is a mouse model for RP 2, Nr2e3 for the Human Enhanced S-cone syndrome (ESCS) 3, Rds for macular dystrophy and RPE65^-/- for LCA 4. Experimental information from the above mentioned models, combined with high-throughput technologies, has led to an increase in the number of experiments related to retinal gene expression.

The recent development of high-throughput technologies has resulted in an enormous volume of gene expression data. General repositories such as GEO 5 and ArrayExpress 6 operate as central data distribution centres encompassing gene expression data from different organisms and from various conditions. In contrast, resources like CGED 7, SIEGE 8 and GeneAtlas 9 are specialized databases that address specific problems; CGED concentrates on gene expression in various human cancer tissues, SIEGE focuses on epithelial gene expression changes induced by smoking in humans and Gene Atlas provides the expression profiles of genes in various mouse and human tissues.

In order to address specific issues related to retina and to meet the needs of retinal biologists in their analysis of gene expression data, we have developed RETINOBASE, a microarray gene expression database for retina. RETINOBASE combines simplified querying, analysis and data visualization options, plus specifically developed meta analysis tools. The integration of gene expression data from various development stages of wild type retina and from diverse conditions and genetic backgrounds will hopefully, not only increase our understanding of the physiological mechanisms involved in normal retinal tissue, but also facilitate studies of gene expression patterns under diverse conditions. Furthermore, RETINOBASE provides a platform for the comparison of different analysis scenarios based on various normalization methods, such as RMA 10, dChip 11, MAS5 12, and clustering methods, such as the K-means 13 and mixture models methods 14.

RETINOBASE uses open-source tools. The website is powered by an Apache web server, PHP and Javascript for dynamic web pages and a PostgreSQL object-relational open source database management system (DBMS) as the back end to store data. The RETINOBASE database schema has been developed using the same philosophy as that used to design BASE 15, with enhancements to accommodate data from different platforms and also complies to the Minimum Information About Microarray Experiment (MIAME) standard 16. It is based on a well-designed relational schema where "realexp" acts as a central table linking expression data with an experiment, sample and array type. This kind of schema helps the system to manage data efficiently, and increases retrieval speed.

RETINOBASE is designed to store gene expression profiles from microarray experiments. We downloaded all publicly available retina-related expression profiles from Gene Expression Omnibus (GEO) yielding 21 experiments 17181920212223242526272829303132, GEO datasets (GSE 1816, 4756, 1835, 3791, 2868). In addition, 8 proprietary experiments have been incorporated that can be accessed with permission from the owner of the experiment. These experiments were performed under different conditions, including knockout models, treatments and time series experiments performed on different organisms such as drosophila, zebra fish, rat, mice and human. All experiments have complete data, except for one experiment 19 that has partial data at the level of fold change, due to the unavailability of raw data (.CEL) or signal intensity data. Currently, RETINOBASE contains approximately 27 million gene expression values resulting from 509 hybridizations. In future releases of the database, we plan to include data from other studies associated with retina, including the SAGE 33, datasets from Diehn and coworkers 34 who used cDNA array to study human eye tissues, and/or datasets from Blackshaw and coworkers 35 who used SAGE to study mouse retinal development.

In order to demonstrate the utility of RETINOBASE, we undertook a case study to identify novel genes that may have a potential role in retinal function. In the experiment "Targeting of GFP to newborn rods by Nrl promoter and temporal expression profiling of flow-sorted photoreceptors" (experiment 7 in RETINOBASE) it was elegantly demonstrated that Nrl (neural retina leucine zipper) is a key regulator of photoreceptor differentiation in mammals 17. We first performed cluster analysis using the "Signal intensity or Cluster query" tool in RETINOBASE by providing Nrl as the gene symbol and then retrieved the resulting clusters. In agreement with the original study by Akimoto et al.,, our "cluster query" found Rho (rhodopsin), Nr2e3 (nuclear receptor subfamily 2, group E, member 3) and Pde6b (phosphodiesterase 6B, cGMP-specific, rod, beta) in the same cluster as Nrl in 4 out of 5 possible combinations (1. RMA normalized data and K-means clustering with TMEV, 2. RMA normalized data, K-means clustering with FASABI, 3. dChip normalized data, K-means clustering with TMEV, 4. dChip normalized data, K-means clustering with FASABI and 5. dChip normalized data, clustering with mixture model), confirming that genes specific for rods are coregulated with Nrl. In addition, Gnat1 (guanine nucleotide binding protein (G protein), alpha transducing activity polypeptide 1), a gene implicated in congenital stationary night blindness 54, was also found in the same cluster in all 5 cluster combinations mentioned above, confirming its role in retinal function. This suggests that Gnat1 is also coregulated with Nrl in retina. Based on the similar coexpression profiles in wild type mouse retina at time points corresponding to embryonic day 16, post natal day 2, 6, 10 and 28 (Figure 5), we further identified a novel gene that is likely to be implicated in regulating retinal differentiation, namely D6Wsu176e (DNA segment, Chr 6, Wayne State University 176, expressed), described as being expressed in the outer nuclear layer of neural retina 55. The RETINOBASE "Ortholog query" for D6Wsu176e points to the human ortholog, FAM3C, that is involved in cell differentiation and proliferation during inner ear embryogenesis 56. With its known function in cell differentiation and its presence in the same cluster as Nrl in 3 out of 5 of the above mentioned clustering combinations, D6Wsu176e may be an interesting candidate for studying rod differentiation. We further went on to check whether Nrl, Rho, Gnat1 and D6Wsu176e genes are coexpressed (present in the same cluster) in other experiments present in RETINOBASE, in particular checking experiment 12 (Gene expression patterns in the retina of rds mice treated with CNTF/rAAV virus and non-treated after 60 days of injection) (GEO: GSE4756) and experiment 14 (Biological characterization of gene response in Rpe65-/- mouse model of Leber's congenital amaurosis during progression of the disease) 21. In these two experiments the four genes mentioned above were present in the same cluster indicating that they might be coregulated. This case study illustrates how RETINOBASE facilitates hypothesis testing for the biologist, and demonstrates how to generate novel hypotheses regarding retinal function and finally, how to identify potential novel targets for human retinopathies.

RETINOBASE is under constant development, including addition of new experiments when available. In addition, data from proprietary experiments can be accessed on approval by individual researchers and will be made generally available after publication. Several functional enhancements are also planned for the future. We will continue to refine and update RETINOBASE with respect to data retrieval, mining and visualization options. Direct upload and meta-analysis options will also be provided.

RETINOBASE has been developed to store, analyse, visualize and compare retinal-related data in order to provide insights into retinal gene expression in various mouse models and other organisms under diverse conditions. Our database, with different types of query options and powerful visualization tools, allows comprehensive analysis of biological mechanisms/pathways of the retina in normal and diseased conditions. We demonstrated by means of a case study how novel genes such as D6Wsu176e (which potentially play an important role in retinal differentiation and development) can be identified using the meta analysis tools incorporated in RETINOBASE. With the addition of new experiments the variety of hypothesis testing options will continuously increase, providing biologists with a valuable tool to gain a better understanding of the retina.

The RETINOBASE can be accessed at 57. All users must register (name and email address) to obtain a username and password.

RK is involved in database design and development, data analysis, design of the user interface and prepared the manuscript. NG, GB and RR developed the web services and database back end. LP is involved in testing various querying tools. WR is involved in data analysis and helped to draft the manuscript. TL participated in the design of the user interface. OP was involved in overall design of the project and in drafting the manuscript.