Skip to main content
Advertisement
Browse Subject Areas
?

Click through the PLOS taxonomy to find articles in your field.

For more information about PLOS Subject Areas, click here.

  • Loading metrics

Using Evolutionary Conserved Modules in Gene Networks as a Strategy to Leverage High Throughput Gene Expression Queries

  • Jeanne M. Serb ,

    serb@iastate.edu

    Affiliation Department of Ecology, Evolution and Organismal Biology, Iowa State University, Ames, Iowa, United States of America

  • Megan C. Orr,

    Affiliation Department of Statistics, Iowa State University, Ames, Iowa, United States of America

  • M. Heather West Greenlee

    Affiliations Bioinformatics and Computational Biology Program, Department of Biomedical Sciences, Iowa State University, Ames, Iowa, United States of America, Bioinformatics and Computational Biology Program, Iowa State University, Ames, Iowa, United States of America

Abstract

Background

Large-scale gene expression studies have not yielded the expected insight into genetic networks that control complex processes. These anticipated discoveries have been limited not by technology, but by a lack of effective strategies to investigate the data in a manageable and meaningful way. Previous work suggests that using a pre-determined seed-network of gene relationships to query large-scale expression datasets is an effective way to generate candidate genes for further study and network expansion or enrichment. Based on the evolutionary conservation of gene relationships, we test the hypothesis that a seed network derived from studies of retinal cell determination in the fly, Drosophila melanogaster, will be an effective way to identify novel candidate genes for their role in mouse retinal development.

Methodology/Principal Findings

Our results demonstrate that a number of gene relationships regulating retinal cell differentiation in the fly are identifiable as pairwise correlations between genes from developing mouse retina. In addition, we demonstrate that our extracted seed-network of correlated mouse genes is an effective tool for querying datasets and provides a context to generate hypotheses. Our query identified 46 genes correlated with our extracted seed-network members. Approximately 54% of these candidates had been previously linked to the developing brain and 33% had been previously linked to the developing retina. Five of six candidate genes investigated further were validated by experiments examining spatial and temporal protein expression in the developing retina.

Conclusions/Significance

We present an effective strategy for pursuing a systems biology approach that utilizes an evolutionary comparative framework between two model organisms, fly and mouse. Future implementation of this strategy will be useful to determine the extent of network conservation, not just gene conservation, between species and will facilitate the use of prior biological knowledge to develop rational systems-based hypotheses.

Introduction

The emergence of system-wide approaches (‘-omics’; e.g., genomics, proteomics, metabolomics, etc.) and related technologies to quantify molecular changes that accompany biological processes or disease states has resulted in an explosion in the amount of data collected by researchers. Investigators across all areas of biology have designed large scale experiments to capture a broader systems-based understanding of gene or protein expression changes that accompany their process of interest. However, many have found that such datasets are too large to be immediately informative, and extracting useful information from these datasets is dependent upon additional analysis.

One strategy to analyze such data is to generate gene network models using one of several analytical frameworks [1][5]. In theory, these network approaches have two advantages: they should accelerate the rate of novel discoveries by automating data analysis and they should be more immune to experimenter bias. This use of computational strategies will potentially lead to discoveries from omics data without a priori knowledge of the system. However, these computational approaches require a tremendous amount of biological data. For example, if an investigator wants to understand which genes function together during a particular developmental process, she might profile changes in gene expression over developmental time. Ideally the number of conditions (e.g., ages, experimental perturbations) under which gene expression is measured should be much larger than the number of genes being profiled in order to obtain an accurate estimate of the covariance matrix upon which the network of all genes is based [6]. Thus, for a microarray experiment that measures the expression of 5000 genes, one should measure the expression of each gene under more than 5000 different conditions. Even collection of 20% of the ideal amount of data for robust analyses is both time and cost prohibitive for most investigators. As a consequence, the majority of biologists collect datasets that are too small for effective computational analysis and too large for systematic and efficient consideration of candidate molecules. This data limbo is a limiting factor to the growth of the field of systems biology.

While it is essential that the development of computational tools and approaches continue, it is also essential that efforts are made to establish ‘biological heuristics’ that will allow benchtop investigators to perform meaningful analyses on the sometimes limited amounts of data they are capable of collecting. A key first step in this process is to consider the development of strategies to efficiently query omics data, as opposed to exhaustively analyzing it. The use of biological heuristics is a flexible strategy, which utilizes prior biological knowledge of the system to design queries. These queries ask specific questions about relatively small groups of interacting genes and return manageable numbers of candidate genes for further analysis at the bench.

Our approach to querying high-throughput data utilizes prior biological knowledge by starting with a ‘seed-network’ of genes, and is based on the paradigm that the expression of genes that function together will change in similar ways over time (i.e., their expression will be correlated). The basic assumption is that if a gene is correlated with one member of the seed network, it may be involved in the process of interest; however, if the same gene is correlated with multiple members of the seed-network it much more likely to be involved in that process (e.g., retinal cell fate determination). One of us has demonstrated previous success identifying gene candidates in development of rod photoreceptors by using a seed-network-based heuristic to query high throughput data [7], and this success motivated our efforts to further develop strategies to identify effective seed networks to query large datasets.

Here we employ our seed-network approach to a genetic comparison of two important models in the study of retinal development: the fly, Drosophila melanogaster, and the mouse, Mus musculus. Despite the morphological and developmental disparity of the fly compound eye [8], [9] and the mouse camera-type eye [10], [11], gene conservation during both fly and mouse retinal development is well-documented [12][16] and there is an implicit assumption of gene regulatory network conservation as well [17], [18]. However the networks are not completely congruent [19]. We test the hypothesis that gene relationships established in the developing fly retina can be identified in correlation networks generated using gene expression data from the developing mouse retina. Further, we hypothesize that the resulting mouse network will be an effective tool to discover candidate genes and gene networks that function during mammalian retinal development. In this report, we take advantage of two biological systems by constructing a ‘comparative seed-network’ based on studies of retinal determination in fly and use it to query gene expression data from the developing mouse retina. Our study was guided by three objectives: 1) to construct a literature-based seed network representing the relationships between genes involved in retinal determination in the fly; 2) to determine whether the network relationships of fly genes are identifiable among homologous mouse genes in expression correlation networks generated from the developing mouse retina; and 3) to assess whether this strategy, based on evolutionary comparison between model organisms, is a useful method to identify biologically relevant candidate genes important in retinal determination. Based on these objectives, our results demonstrate successful application of this strategy within our experimental system and provide a clear framework to evaluate this approach in other biological areas.

Results

Seed network construction in fly

Seed networks are graphs that represent relationships among genes during a biological process, such as retinal determination. These relationships may be physical interactions or causal relations by direct or indirect means, and are represented as edges in the graph or connections (links) in the gene network. We used the results of published experimental studies on eye differentiation in fly to identify a set of 18 genes implicated in fly retinal development, which was built off of the fly retinal determination gene network (RDGN) [12]. We integrated these data into a comprehensive fly ‘seed network’ (Figure 1) based on the work described in File S1.

thumbnail
Figure 1. The fly seed network based on experimental results in the literature.

Positive correlations between genes are represented by blue edges and negative correlations are represented by red edges. Full names and their abbreviations for Drosophila genes are provided in Table 1.

https://doi.org/10.1371/journal.pone.0012525.g001

Extraction of homologous seed network from mouse datasets

To determine whether the gene relationships represented in the fly seed network are represented in the developing mouse retina, we first converted the literature-based fly seed network into a mouse gene network of putative homologs (Table 1). Then we used BioNet Workbench [http://bionetworkbench.sourceforge.net/] to query four previously published gene expression datasets (I–IV) from mouse [20][23]. The datasets were queried for pairwise correlations of >|0.65| between all mouse genes that are homologous to fly seed network members (“seed genes”). A summary of the seed genes and their pairwise correlation values (if >|0.65|) in each of the mouse datasets are given in Table 2. The result was a mouse seed network “extracted” from published gene expression datasets for mouse.

thumbnail
Table 1. Fly genes from the seed network and their putative mouse homologs.

https://doi.org/10.1371/journal.pone.0012525.t001

thumbnail
Table 2. Correlation of network seed genes in each of the four expression datasets of mouse.

https://doi.org/10.1371/journal.pone.0012525.t002

Based on the finding that a subset of relationships from the fly seed network appear to be conserved in the developing mouse retina, we hypothesized that the extracted seed-network (ESN) of mouse gene relationships would be useful for querying the mouse gene expression data to identify additional candidate gene network members. To identify candidates, genes correlated >|0.65| with each gene in the extracted seed-network were retrieved from each dataset and the lists were compiled. Lists of genes correlated with each ESN gene were analyzed to identify genes that correlated with more than one gene in the ESN. Based on the paradigm that genes correlated with multiple ESN genes are likely to have a functional relationship to the gene network, we focused our analysis on 46 candidate genes that were correlated with three or more ESN members (Table 3). Among these 46, 39 genes were correlated minimally with Eya1, Notch1 and Six3. We evaluated the relevance of candidate genes identified by this comparative seed-network approach in three ways.

thumbnail
Table 3. List of 46 candidate genes found in three or more seed gene lists.

https://doi.org/10.1371/journal.pone.0012525.t003

First, we performed a manual literature search to find reports of candidate genes' association with the retina, retinal development, brain development or other developmental processes. Results from this manual search are given in Table 3. Forty out of 46 (86%) candidate genes have been previously reported to be associated with one or more of these topics [20], [24][86]. Additionally, eight candidate genes (17%) are associated with retinal ganglion cells (RGCs) or RGC development in previous experimental studies [26], [30], [36], [48], [53], [55], [58], [62], [64], [67], [73], [78], [81] (Table 3).

Second, we examined the spatial and temporal expression of six candidate genes in the developing mouse retina. We chose to examine candidates that had been previously reported to be associated with the developing brain, but not the developing retina, and had commercially available antibodies. Using immunohistochemistry in retinal tissue sections from mice ages embryonic day (E)13, E15, E17 and postnatal day (P)0, P5 and P10, we characterized the expression of APLP2, DPYS14, NDN, PAFAH1B3, PSME1 and TMSB10. Candidates were considered highly relevant if they were: 1) expressed in the developing retina, 2) exhibited specific (as opposed to diffuse) localization in the developing retina, and/or 3) the localization of the immunoreactivity changed as the retina matured. Based on these criteria, five of the six candidate genes tested that had not been previously associated with retinal development (aplp2, ndn, pafah1b3, psme-1 and tsmb10) were considered good candidates for further investigation (Figures 2, 3, 4, 5 and 6).

thumbnail
Figure 2. Dynamic protein expression of APLP2 in developing mouse retina.

APLP2-IR in the E13 mouse retina was slightly more intense in cells in the inner and outermost retina (A). In the E15 mouse retina (B), APLP2-IR was observed throughout the thickness of the retina, though the most intense immunoreactivity remained localized to cells in the in the inner and outermost retina. By E17, APLP2-IR was largely restricted to cells and processes in the inner one-third of the retina (C). By the day of birth (P0), APLP2-IR was restricted to cell bodies in the ganglion cell layer (GCL), the IPL and the inner nuclear layer (INL). In the P5 retina, APLP2-IR was most prominent in the IPL, GCL and OFL, though some punctate APLP2-IR remained in the INL (D). By P10, APLP2-IR was further restricted to the IPL and OFL, with punctate immunoreactivity only present in the GCL. Abbreviations: GCL, ganglion cell layer; INL, inner nuclear layer; IPL, inner plexiform layer; NBL, neuroblastic layer; OFL, optic fiber layer; ONL, outer nuclear layer; RPE, retinal pigment epithelium. Bars, 30 µm.

https://doi.org/10.1371/journal.pone.0012525.g002

thumbnail
Figure 3. Dynamic protein expression of NDN in developing mouse retina.

NDN-IR in the E13 mouse retina was localized to cells in the inner one-third of the retina (A). Similarly, in the E15 (B) and E17 (C) retinas NDN-IR was observed in the inner retina, including the GCL and OFL. In the P0 retina, NDN-IR was restricted to the developing IPL and OFL (D). Similarly, in the P5 and P10 retinas, NDN-IR remained in the IPL and OFL, respectively (E and F). Abbreviations same as in Figure 2. Bars, 30 µm.

https://doi.org/10.1371/journal.pone.0012525.g003

thumbnail
Figure 4. Dynamic protein expression of PAFAH1B3 in developing mouse retina.

In the E13, E15 and E17 mouse retinas, PAFAH1B3-IR was observed throughout the thickness of the retina, though was slightly more intense in the cells of the inner retina (A–C). However, in the P0 retina, PAFAH1B3-IR was restricted to the GCL and OFL (D). PAFAH1B3-IR in the P5 retina was further restricted to a subset of cells in the GCL and the OFL (E). Pafah1b3-IR in the P10 retina was decreased to a punctate pattern in the GCL (F). Abbreviations same as in Figure 2. Bars, 30 µm.

https://doi.org/10.1371/journal.pone.0012525.g004

thumbnail
Figure 5. Dynamic protein expression of Psme1 in developing mouse retina.

PSME1-IR in the E13 mouse retina was diffusely distributed throughout the E13, E15 and E17 retinas (A–C). However, by P0, PSME1-IR was no loner detectable above background (D). Similarly, no PSME-IR was detected in the P5 or P10 retinas (E, F). Abbreviations same as in Figure 2. Bars, 30 µm.

https://doi.org/10.1371/journal.pone.0012525.g005

thumbnail
Figure 6. Dynamic protein expression of TSMB10 in developing mouse retina.

TSMB10-IR in the E13 and E15 mouse retina was distributed throughout the retina (A, B). In the E17 mouse retina, TSMB10-IR was more intense in the inner one-third of the retina (C). By P0 TSMB10-IR in the mouse retina was largely restricted to the IPL, GCL and OFL (D). Similarly in the P5 and P10 retinas, TSMB10-IR was observed in the IPL, GCL and OFL (E, F). Abbreviations same as in Figure 2. Bars, 30 µm.

https://doi.org/10.1371/journal.pone.0012525.g006

Third, we used the biological process GO annotations Nervous System Development (0007399) and Neural Retina Development (0003407), to statistically evaluate our candidate list. In the list of 46 candidates identified by using our seed-network approach, seven of the genes had a Nervous System Development GO annotation. By using a Fisher's exact test we determined that Nervous System Development is over-represented among the group of candidate genes. The p-value for this test was 0.026, which represents the probability of seeing seven or more Nervous System Development genes in a list of 46 genes randomly selected from the 8544 genes represented on the Murine Genome U74Av2 array. Because it would be unlikely to see so many Nervous System Development genes in our candidate list of 46 genes by chance, our results suggest that Nervous System Development genes were overrepresented in our candidate list.

In summary, our analysis identified a network of 46 highly correlated candidate genes. Expression of 22 (∼47%) of these candidate genes has been previously reported in the retina or developing retina (see references in Table 3), although their specific relationship to genes within the retinal determination gene network has not been reported. We examined six candidate genes that had previously been associated with brain development, and determined that five of these genes have dynamic spatial and temporal expression in the developing mouse retina. Finally, of these 46 mammalian genes, 42 (∼91%) have homologs in fly, making them potential candidates for studies of fly retinal development as well. These findings demonstrate the powerful advantages of integrating evolutionary comparisons and systems approaches, even when approaching well-studied biological questions.

Discussion

The compound eye of Drosophila is an outstanding model system to study the molecular basis of eye specification, in part, because retinal development is an organized, step-wise process with clearly demarcated regions of cell differentiation and patterning [8], [87]. These properties of the fly model have facilitated the elucidation of genetic networks involved in retinal cell differentiation and the identification of key genes required for retinal development in fly. Comparative studies between model organisms [12], [18] led to discoveries that homologous genes play important and similar roles in fly and mammalian retinal development and many of these key genes have similar connectivity in gene networks [19]. This principle of gene network conservation has motivated our development of the seed-network strategy, which we have presented here, and provides a way to validate our novel heuristic approach.

We tested our strategy using gene expression datasets from the developing mouse retina. The results from this study support our hypothesis that gene relationships in the developing fly retina are identifiable in correlation networks generated using gene expression data from the developing mouse retina. While not all gene relationships in the fly network were identified in the mouse ESN, this is not unexpected. Our results provide support for the assumption that there will be a degree of conservation within genetic networks of homologous genes, even between highly divergent species such as fly and mouse. Complete congruence between the RDGN of fly and mouse would be surprising given that these organisms possess highly divergent eye morphologies. Our results also support the second hypothesis that the mouse network derived from relationships between homologous genes from the fly RDGN (i.e. our extracted seed network [ESN]), would be an effective way to discover high quality candidate genes involved in retinal development in mouse. Our queries identified a reasonable number (46) of candidates, when compared to the hundreds or thousands of genes that correlate with a single gene of interest. The majority of our candidate genes were correlated (positively or negatively) with the same three seed genes (Notch1, Eya1 and Six3) suggesting that these three seed genes are at the functional core of this network regulating retinal development in mouse.

At the heart of our approach is the development of biological heuristics to focus queries of relatively sparse (albeit typical) expression datasets from the developing mouse retina. It is important to note that this approach is intended to facilitate the formulation of hypotheses by providing a mechanism to integrate prior biological knowledge, but not intended to make conclusions about the function or assign significance to the candidates. The use of relationships among genes as a biological heuristic to query high-throughput data, as opposed to queries based on single genes, appears to be more profitable and efficient for the identification of additional candidates. Thus, the candidate genes we identified in this study are not end points, but are the basis of hypotheses to guide future experimental work. Traditional wet-lab experiments will be required to test these hypotheses of the specific role of each candidate gene and its placement in the gene regulatory network during mouse retinal development.

From a comparative evolutionary perspective, our results underscore the importance of looking for conservation of networks, and not just conservation at the level of individual genes. While gene orthologs may function in a similar way in a complex process or a disease state in different organisms, it is the conservation of not only the gene, but of its relationships to other genes in a network, that dramatically increases the likelihood that the gene, in fact, functions similarly. Although it has been directly demonstrated in only a few cases [19], [88][91], regulatory network conservation has long been the rationale for the use of model organisms to study human diseases. Comparative studies that investigate the extent of conservation in developmental regulatory networks (and of characteristics, such as modularity, connectivity, etc.) are beginning to identify common themes in networks that direct organogenesis, e.g., [92]. While it is unreasonable to expect that genetic regulatory networks controlling the development of organs in highly divergent organisms will be conserved in their entirety, application of the approach proposed here to identify conserved network modules should allow systems biologists to better capitalize on what is known in one species to advance discovery in another.

Materials and Methods

Our biological heuristic strategy is described below and summarized in Figure 7.

thumbnail
Figure 7. A description of our biological heuristic strategy.

https://doi.org/10.1371/journal.pone.0012525.g007

Construction of the fly seed network

We identified 18 genes in Drosophila that are involved in the retinal determination gene network (RDGN), based on published literature. Previous researchers have identified individual relationships of these genes to one another and these experimentally-determined relationships among the 18 genes were the basis of our fly seed network (see descriptions and citations in File S1).

Homolog identification of seed network genes

Homolog identification can be difficult when comparing genomes across great evolutionary time as a result of sequence evolution and paralogous duplication events within a lineage. Because of these issues, we identified putative mouse orthologs of the Drosophila seed network manually, using a combination of approaches, including examination of the genomic databases FlyBase [http://flybase.org/] [93] and Mouse Genome Informatics [MGI; http://www.informatics.jax.org/] [94], phylogenetic methods presented in TreeFam [http://www.treefam.org/] [95], [96], and HomoloGene [http://www.ncbi.nlm.nih.gov/homologene] [97]. Additional assignment of orthology between fly and mouse genes was based on experimental data. For example, the mouse has three Teashirt (tsh)-like genes, Tshz1, Tshz2 and Tshz3, all of which can rescue tsh null mutants and induce ectopic eyes in the fly [98]. Likewise, we designated the Pax6 isoform, Pax6(5a), found in humans and mouse, as the ortholog for fly eyg because the genes are structurally related [99], and we treated the mouse gene Math5 (Atonal7) as the homolog to the fly gene Atonal based on others' work [100], [101] reviewed in [19]. Finally, qualitative and functional comparisons of the mouse genes Six3 and Six6 to so and optix in fly, suggest that optix should be treated as an ortholog of Six3 and Six6 [102], reviewed in [19]. Table 1 lists fly seed network genes and their mouse homolog assignment based on these data.

Description of mouse data sets and construction of ESN

Published, freely available datasets measuring gene or protein expression in the developing mouse retina at multiple time points were collected and preprocessed as described in Hecker et al. [7]. Each dataset represents expression data collected from developing mouse retinae at multiple time points and includes: a SAGE (serial analysis of gene expression) of whole retina from Blackshaw et al. [20] was downloaded from online supplementary material; one cDNA microarray of whole retina from Zhang et al. [21] was downloaded from online supplementary material; and two Affymetrix microarrays of whole retina, the Mu74Av2 chip from Liu et al. [22] was downloaded from GEO (GDS 1845) and the Mu74Av2_1 chip from Dorrell et al. [23] was downloaded from http://www.scripps.edu/cb/friedlander/gene_expression/). These mouse datasets were designated as I, II, III, and IV, respectively, and were saved in BioNet Workbench [http://bionetworkbench.sourceforge.net/] for analysis.

We calculated Spearman Rank pairwise correlations in each mouse expression dataset using BioNet Workbench to construct the extracted seed network (ESN) for mouse. Correlation networks provide a visual representation of pairwise associations between genes in large data sets consisting of expression measurements for hundreds or thousands of genes. In a gene or protein expression correlation network, the nodes represent the genes or proteins and weighted links model interactions between them. The weight associated with a link between a pair of nodes models the correlation estimated from measurements of expression (e.g., mRNA or protein) levels of the corresponding genes across a set of experimental conditions or time points. The Spearman rank correlation measure, which assumes only an arbitrary monotonic, not necessarily linear, relationship between variables being correlated, has been demonstrated to be effective for detecting functional relationships between genes [103]. Correlation coefficients using time-course expression data are calculated by a measure of how the expression levels between any given pair of genes changes over time. Genes that are perfectly correlated with one another have a correlation coefficient of 1. Gene pairs whose expression is exactly the opposite of one another have a correlation coefficient of −1. Two genes whose expression is not correlated (no different than random) have a correlation coefficient of 0. In cases where multiple mouse paralogs for a single fly gene are present, each paralog was queried separately. Not all mouse seed genes were present in all datasets.

A link between a pair of seed network genes is supported by a dataset if the corresponding genes are positively or negatively correlated in that dataset, with the absolute value of correlation greater than or equal to 0.65 in one of the mouse datasets (I–IV). Our choice of the threshold of 0.65 for correlation was influenced by similar choices in previous studies [104][106] that have revealed biologically relevant links between co-expressed genes. It should be noted that we do not assign statistical significance to the value of the correlation coefficient, but rather consider the value a flexible tool that can be used at the discretion of the investigator to filter gene lists to a manageable number of candidates. This is appropriate, as our strategy is designed to facilitate the generation of hypotheses, not conclusions.

Generating candidate gene lists

To identify candidate genes that may be involved in the gene network controlling mouse development, we used genes from the extracted seed-network (ESN; mouse homologs of fly RDGN genes whose pairwise expression correlation coefficients were >|0.65| in at least one dataset) to query large-scale gene expression datasets of the developing retina (I–IV). Each of the 17 genes from the ESN (a.k.a. seed genes) in mouse was examined separately in datasets I–IV to develop candidate gene lists. Lists were compiled by identifying genes that were correlated with individual seed genes, with a correlation coefficient greater than |0.65|. Then, gene lists for all seed genes were compared to identify candidate genes that correlated with more than one seed gene. Genes that correlated with more than three seed genes were investigated for potential biological relevance in mouse retinal development.

Biological relevance of candidate genes was assessed using manual PubMed [http://www.ncbi.nlm.nih.gov/pubmed/] [107] searches with the following search terms: retina, retinal development, CNS development, brain development, development. Gene synonymies used in the literature searches of the candidate mouse genes are listed in Table 3. Putative Drosophila homologs of the mouse genes from the candidate gene list were identified using FlyBase [http://flybase.org/] [93], the Mouse Genome Database (MGD) [URL: http://www.informatics.jax.org; January, 2009] [108], and TreeFam [http://www.treefam.org/] [95], [96]. When necessary, paralogous genes were included to more fully capture gene homology between the two organismal models.

In order to determine if the GO annotation Nervous System Development (0007399) was over-represented among the 46 candidate genes, a Fisher's exact test was used. Over-representation was declared if the number of genes with the GO annotation of interest on our candidate list was significantly higher than would be expected by chance, i.e., if the observed number of Nervous System Development genes was greater than would be expected when randomly selecting 46 genes from a collection of 550 Nervous System Development genes mixed with 7994 other genes. Information from version 2.4.1 of the R statistical software annotation package mgu74av2.db [http://www.bioconductor.org/] and release 30 of the NetAffx annotation file for the Murine Genome U74Av2 array [http://www.affymetrix.com/] was combined in order to perform the Fisher's exact test. Due to relatively frequent changes in probe set annotations and gene symbols, and also due to slight disagreements between the two annotation sources, a conservative analysis was performed. Although there are likely more genes represented on the Murine Genome U74Av2, the total number of genes represented was declared to be 8544. Similarly, the number of genes with Nervous System Development annotation was declared to be 550, although this number is likely high. Using an under-estimate of the total number of genes and an over-estimate of the genes with Nervous System Development annotation results in a higher p-value and thus more conservative results than if the true values were used.

It should also be noted that no probe set was mapped to the gene Prrt1 in either of the current annotation sources (although it was in previous annotations), but this gene was included in the analysis because of its inclusion in the analysis of previous papers. Removing this gene from the analysis does not affect the significance of our results.

Examining candidate gene expression with immunohistochemistry

To investigate the biological relevance of the candidate genes correlated with the ESN, we examined the spatial and temporal expression of six candidate genes in the developing mouse retina. Tissue was prepared from C57BL/6 mice in a colony maintained at Iowa State University. The gestational period of C57BL/6 mice is approximately 19 days and date of birth is designated as postnatal day 0 (P0). The developmental time series investigated included pups from embryonic days 13, 15 and 17 and postnatal days 0, 5, and 10. Mice were euthanized and their heads were removed and immersion fixed in 4% paraformaldehyde in 0.1M PO4 buffer (pH 7.5). The tissue was cryoprotected in a 30% sucrose solution in 0.1 M PO4 buffer (pH 7.4) and embedded in OCT mounting media. Tissue was sectioned at a thickness of 20 µm on a cryostat and the sections were thaw-mounted onto microscope slides and stored at −20°C. All animal procedures had the approval of the ISU committee on animal care.

Frozen tissue sections were rinsed in 0.5M KPBS and incubated in blocking solution consisting of KPBS containing 1% bovine serum albumin (Fisher, Pittsburgh, PA), 0.4% Triton-X 100 (Sigma), and 1.5% normal donkey serum (Invitrogen) for 2 hours. Cells were incubated in primary antibody overnight at 4°C. The following day slides (tissue or cells) were washed in KPBS containing 0.02% Triton-X 100 after which fluorescent secondary antibody was applied for 2 hours. After washes in KPBS containing 0.02% Triton-X 100, the slides were incubated in 300 µM DAPI diluted in KPBS. The slides were rinsed in KPBS before cover-slipping with Vectashield fluorescence mounting medium (Vector Laboratories, Burlingame, CA). The antibody sources and concentrations used for the immunohistochemical analysis are summarized in Table S1.

Supporting Information

File S1.

Seed network construction in fly.

https://doi.org/10.1371/journal.pone.0012525.s001

(0.12 MB DOC)

Table S1.

The antibody sources and concentrations used for the immunohistochemical analysis.

https://doi.org/10.1371/journal.pone.0012525.s002

(0.04 MB DOC)

Acknowledgments

The authors would like to thank Xia Zhang for data organization, Fadi Towfic and Nada Pavlovic for technical support, and Dan Nettleton for helpful discussion. We also thank Drena Dobbs, Donald Sakaguchi and members of the J.M. Serb, Dean Adams, and Nicole Valenzuela labs for constructive reading of the manuscript.

Author Contributions

Conceived and designed the experiments: JMS MHWG. Performed the experiments: JMS MHWG. Analyzed the data: JMS MCO MHWG. Wrote the paper: JMS MCO MHWG.

References

  1. 1. Hartemink AJ, Gifford DK, Jaakkola TS, Young RA (2002) Combining location and expression data for principled discovery of genetic regulatory network models. Pacific Symposium on Biocomputing 2002: 437–449.
  2. 2. Markowetz F, Spang R (2007) Inferring cellular networks–a review. BMC Bioinformatics 8: Suppl 6S5.
  3. 3. Magwene PM, Kim J (2004) Estimating genomic coexpression networks using first-order conditional independence. Genome Biology 5: R100.
  4. 4. Butte A, Kohane I (2003) Relevance networks: a first step towards finding genetic regulatory networks within microarray data. In: Parmigiani G, Garret ES, Irizarry RA, Zeger SL, editors. The analysis of gene expression data. New York: Springer.
  5. 5. Friedman N, Linial M, Nachman I, Pe'er D (2000) Using Bayesian networks to analyze expression data. Journal of Computing Biology 7: 601–620.
  6. 6. Johnson RA, Wichern DW (2002) Applied multivariate statistical analysis. Prentice Hall.
  7. 7. Hecker LA, Alcon TA, Honavar VG, West Greenlee MH (2008) Using a seed-network to query multiple large-scale gene expression datasets from the developing retina in order to identify and prioritize experimental targets. Bioinformatics and Biology Insights 2008: 91–102.
  8. 8. Pappu KS, Mardon G (2004) Genetic control of retinal specification and determination in Drosophila. International Journal of Developmental Biology 48: 913–924 SI.
  9. 9. Land MF, Nilsson D-E (2006) Animal Eyes. Oxford University Press.
  10. 10. Livesey FJ, Cepko CL (2001) Vertebrate neural cell-fate determination: lessons from the retina. Nature Reviews 2: 109–118.
  11. 11. Walls GL (1967) The vertebrate eye and its adaptive radiation. New York: Haffner Publishing Company. 785 p.
  12. 12. Quiring R, Walldorf U, Kloter U, Gehring WJ (1994) Homology of the eyeless gene of Drosophila to the small eye gene in mice and aniridia in humans. Science 265: 785–789.
  13. 13. Xu P, Woo I, Her HB, D.R., Maas RL (1997) Mouse Eya homologues of the Drosophila eyes absent gene require Pax6 for expression in lens and nasal placodes. Development 124: 219–231.
  14. 14. Brown NL, Kanekar S, Vetter ML, Tucker PK, Gemza DL, et al. (1998) Math5 encodes a murine basic helix-loop-helix transcription factor expressed during early stages of retinal neurogenesis. Development 125: 4821–4833.
  15. 15. Niiya A, Ohto H, Kawakami K, Araki M (1998) Localization of Six4/AREC3 in the developing mouse retina; implications in mammalian retinal development. Experimental Eye Research 67: 699–707.
  16. 16. Chen R, Halder G, Zhang Z, Mardon G (1999) Signaling by the TGF-beta homolog decapentaplegic functions reiteratvely within the network of genes controlling retinal cell fate determination in Drosophila. Development 126: 935–943.
  17. 17. Wawersik S, Maas RL (2000) Vertebrate eye development as modeled in Drosophila. Human Molecular Genetics 9: 917–925.
  18. 18. Gehring WJ, Ikeo K (1999) Pax6: mastering eye morphogenesis and eye evolution. TIG 15: 371–377.
  19. 19. Donner AL, Maas RL (2004) Conservation and non-conservation of genetic pathways in eye specificiation. Int J Dev Biol 48: 743–753.
  20. 20. Blackshaw S, Harpavat S, Trimarchi J, Cai L, Huang H, et al. (2004) Genomic analysis of mouse retinal development. PLoS Biology 2:
  21. 21. Zhang SS, Xu X, Liu MG, Zhao H, Soares MB, et al. (2006) A biphasic pattern of gene expression during mouse retina development. BMC Developmental Biology 6: 48.
  22. 22. Liu J, Wang J, Huang Q, Higdon J, Magdaleno S, et al. (2006) Gene expression profiles of mouse retinas during the second and third postnatal weeks. Brain Research 1098: 113–125.
  23. 23. Dorrell MI, Aguilar E, Weber C, Friedlander M (2004) Global gene expression analysis of the developing postnatal mouse retina. Investigative Ophthalmology and Visual Science 45: 1009–1019.
  24. 24. Ahmad H, Singh SV, Medh RD, Ansari GA, Kurosky A, et al. (1988) Differential expression of alpha, mu and pi classes of isozymes of glutathione S-transferase in bovine lens, cornea, and retina. Archives of Biochemistry and Biophysics 266: 416–426.
  25. 25. Ahuja P, Caffé AR, Ahuja S, Ekström P, van Veen T (2005) Decreased glutathione transferase levels in rd1/rd1 mouse retina: replenishment protects photoreceptors in retinal explants. Neuroscience 131: 935–943.
  26. 26. Allen GW, Liu J, Kirby MA, De León M (2001) Induction and axonal localization of epitheliałepidermal fatty acid-binding protein in retinal ganglion cells are associated with axon development and regeneration. Journal of Neuroscience Research 66: 396–405.
  27. 27. Amrein L, Barraud P, Daniel JY, Perel Y, Landry M (2005) Expression patterns of nm23 genes during mouse organogenesis. Cell Tissue Research 322: 365–378.
  28. 28. Asano T, Shinohara H, Morishita R, Ueda H, Kawamura N, et al. (2001) Selective localization of G protein gamma5 subunit in the subventricular zone of the lateral ventricle and rostral migratory stream of the adult rat brain. Journal of Neurochemistry 79: 1129–1135.
  29. 29. Azuma M, Shearer TR (2008) The role of calcium-activated protease calpain in experimental retinal pathology. Survey of Opthalmology 53: 150–163.
  30. 30. Birgbauer E, Cowan CA, Sretavan DW, Henkemmeyer M (2000) Kinase independent function of EphB receptors in retinal axon pathfinding to the optic disc from dorsal but not ventral retina. Development 127: 1231–1241.
  31. 31. Border BG, Lin SC, Griffin WS, Pardue S, Morrison-Bogorad M (1993) Alterations in actin-binding beta-thymosin expression accompany neuronal differentiation and migration in rat cerebellum. Journal of Neurochemistry 61: 2104–2014.
  32. 32. Bult C, Eppig JT, Richardson JE, Blake JA, Group atmotMGD (2008) The Mouse Genome Database (MGD): mouse biology and model systems. Nucleic Acids Research 36: D724–728.
  33. 33. Carpintero P, Anadon R, Gomez-Marquez J (1999) Expression of the thymosin beta10 gene in normal and kainic acid-treated rat forebrain. Brain Research Mol Brain Res 18: 141–146.
  34. 34. Charrier E, Mosinger B, Meissirel C, Aguera M, Rogemond V, et al. (2006) Transient alterations in granule cell proliferation, apoptosis and migration in postnatal developing cerebellum of CRMP1-/- mice. Genes to Cells 11: 1337–1352.
  35. 35. Chen Y, Tang BL (2006) The amyloid precursor protein and postnatal neurogenesis/neuroregeneration. Biochemical and Biophysical Research Communications 341: 1–5.
  36. 36. Corbetta S, Gualdoni S, Ciceri G, Monari M, Zuccaro E, et al. (2009) Essential role of Rac1 and Rac3 GTPases in neuronal development. FASEB J 23: 1347–1357.
  37. 37. Davis JA, Reed RR (1996) Role of Olf-1 and Pax-6 transcription factors in neurodevelopment. Journal of Neuroscience 16: 5082–5094.
  38. 38. de Brouwer AP, Williams KL, Duley JA, van Kuilenburg AB, Nabuurs SB, et al. (2007) Arts syndrome is caused by loss-of-function mutations in PRPS1. American Journal of Human Genetics 81: 507–518.
  39. 39. Dyer MA, Cepko CL (2001) p27Kip1 and p57Kip2 regulate proliferation in distinct retinal progenitor cell populations. Journal of Neuroscience 21: 4259–4271.
  40. 40. Dyer MA, Cepko CL (2001) The p57Kip2 cyclin kinase inhibitor is expressed by a restricted set of amacrine cells in the rodent retina. Journal of Comparative Neurology 429: 601–614.
  41. 41. El-Khodor BF, Kholodilov NG, Yarygina O, Burke RE (2001) The expression of mRNAs for the proteasome complex is developmentally regulated in the rat mesencephalon. Brain Research, Developmental Brain Research 129: 47–56.
  42. 42. Ewing JF, Maines MD (1995) Immunohistochemical localization of biliverdin reductase in rat brain: age related expression of protein and transcript. Brain Research 672: 29–41.
  43. 43. Fullerton MD, Hakimuddin F, Bakovic M (2007) Developmental and metabolic effects of disruption of the mouse CTP:phosphoethanolamine cytidylyltransferase gene (Pcyt2). Molecular Cellular Biology 27: 3327–3336.
  44. 44. Gu C, Zhou W, Puthenveedu MA, Xu M, Jan YN, et al. (2006) The microtubule plus-end tracking protein EB1 is required for Kv1 voltage-gated K+ channel axonal targeting. Neuron 52: 803–816.
  45. 45. Guénette S, Chang Y, Hiesberger T, Richardson JA, Eckman CB, et al. (2006) Essential roles for the FE65 amyloid precursor protein-interacting proteins in brain development. EMBO Journal 25: 420–431.
  46. 46. Hou ST, Jiang SX, Desbois A, Huang D, Kelly J, et al. (2006) Calpain-cleaved collapsin response mediator protein-3 induces neuronal death after glutamate toxicity and cerebral ischemia. Journal of Neuroscience 26: 2241–2209.
  47. 47. Hwang IK, Yoo KY, Kim DW, Kim SY, Park JH, et al. (2008) Ischemia-induced ribosomal protein S3 expressional changes and the neuroprotective effect against experimental cerebral ischemic damage. Journal of Neuroscience Research 86: 1823–1835.
  48. 48. Ivanov D, Dvoriantchikova G, Barakat DJ, Nathanson L, Shestopalov VI (2008) Differential gene expression profiling of large and small retinal ganglion cells. Journal of Neuroscience Methods 174: 10–17.
  49. 49. Jones SE, Jomary C, Grist J, Stewart HJ, Neal MJ (2000) Identification by array screening of altered nm23-M2/PuF mRNA expression in mouse retinal degeneration. Molecular Cell Biol Res Commun 4: 20–25.
  50. 50. Kanan Y, Moiseyev G, Agarwal N, Ma JX, Al-Ubaidi MR (2007) Light induces programmed cell death by activating multiple independent proteases in a cone photoreceptor cell line. Investigative Ophthalmology and Visual Science 48: 40–51.
  51. 51. Kasai T, Inoue M, Koshiba S, Yabuki T, Aoki M, et al. (2004) Solution structure of a BolA-like protein from Mus musculus. Protein Science 13: 545–548.
  52. 52. Kim HJ, Sohn KM, Shy ME, Krajewski KM, Hwang M, et al. (2007) Mutations in PRPS1, which encodes the phosphoribosyl pyrophosphate synthetase enzyme critical for nucleotide biosynthesis, cause hereditary peripheral neuropathy with hearing loss and optic neuropathy (cmtx5). American Journal of Human Genetics 81: 552–558.
  53. 53. Koyabu Y, Nakata K, Mizugishi K, Aruga J, Mikoshiba K (2001) Physical and functional interactions between Zic and Gli proteins. Journal of Biological Chemistry 276: 6889–6892.
  54. 54. Lai M, Wang F, Rohan JG, Maeno-Hikichi Y, Chen Y, et al. (2005) A tctex1-Ca2+ channel complex for selective surface expression of Ca2+ channels in neurons. Nature Neuroscience 8: 435–442.
  55. 55. Lee R, Petros TJ, Mason CA (2008) Zic2 regulates retinal ganglion cell axon avoidance of ephrinB2 through inducing expression of the guidance receptor EphB1. Journal of Neuroscience 28: 5910–5919.
  56. 56. Lee SS, Seo HS, Choi SJ, Park HS, Lee JY, et al. (2003) Characterization of the two genes differentially expressed during development in human fetal astrocytes. Yonsei Medical Journal 44: 1059–1068.
  57. 57. Lim YS, McLaughlin T, Sung TC, Santiago A, Lee KF, et al. (2008) p75(NTR) mediates ephrin-A reverse signaling required for axon repulsion and mapping. Neuron 59: 746–758.
  58. 58. Liu JW, Almaguel FG, Bu L, De Leon DD, De Leon M (2008) Expression of E-FABP in PC12 cells increases neurite extension during differentiation: involvement of n-3 and n-6 fatty acids. Journal of Neurochemistry 106: 2015–2029.
  59. 59. Lobo MK, Yeh C, Yang XW (2008) Pivotal role of early B-cell factor 1 in development of striatonigral medium spiny neurons in the matrix compartment. Journal of Neuroscience Research 86: 2134–2146.
  60. 60. McKee AE, Minet E, Stern C, Riahi S, Stiles CD, et al. (2005) A genome-wide in situ hybridization map of RNA-binding proteins reveals anatomically restricted expression in the developing mouse brain. BMC Developmental Biology 5: 14.
  61. 61. Miller NL, Wevrick R, Mellon PL (2009) Necdin, a Prader-Willi syndrome candidate gene, regulates gonadotropin-releasing hormone neurons during development. Human Molecular Genetics 18: 248–260.
  62. 62. Mitchell DC, Bryan BA, Liu JP, Liu WB, Zhang L, et al. (2007) Developmental expression of three small GTPases in the mouse eye. Molecular Vision 13: 1144–1153.
  63. 63. Monaghan MM, Trimmer JS, Rhodes KJ (2001) Experimental localization of Kv1 family voltage-gated K+ channel alpha and beta subunits in rat hippocampal formation. Journal of Neuroscience 21: 5973–5783.
  64. 64. Munemasa Y, Kim SH, Ahn JH, Kwong JM, Caprioli J, et al. (2008) Protective effect of thioredoxins 1 and 2 in retinal ganglion cells after optic nerve transection and oxidative stress. Investigative Ophthalmology and Visual Science 49: 3535–3543.
  65. 65. Nagahama H, Hatakeyama S, Nakayama K, Nagata M, Tomita K, et al. (2001) Spatial and temporal expression patterns of the cyclin-dependent kinase (CDK) inhibitors p27Kip1 and p57Kip2 during mouse development. Anatomy and Embryology 203: 77–87.
  66. 66. Nakajima E, David LL, Bystrom C, Shearer TR, Azuma M (2006) Calpain-specific proteolysis in primate retina: Contribution of calpains in cell death. Investigative Ophthalmology and Visual Science 47: 5469–5475.
  67. 67. Nakazawa T, Nakano I, Furuyama T, Morii H, Tamai M, et al. (2000) The SCG10-related gene family in the developing rat retina: persistent expression of SCLIP and stathmin in mature ganglion cell layer. Brain Research 861: 399–407.
  68. 68. Okazaki Y, Furuno M, Kasukawa T, Acachi J, Team. FCRGERGPII (2002) Analysis of the mouse transcriptome based on functional annotation of 60,770 full-length cDNAs. Nature 420: 563–573.
  69. 69. Panahian N, Huang T, Maines MD (1999) Enhanced neuronal expression of the oxidoreductase–biliverdin reductase–after permanent focal cerebral ischemia. Brain Research 850: 1–13.
  70. 70. Quach TT, Massicotte G, Belin MF, Honnorat J, Glasper ER, et al. (2008) CRMP3 is required for hippocampal CA1 dendritic organization and plasticity. FASEB J 22: 401–409.
  71. 71. Quach TT, Mosinger BJ, Ricard D, Copeland NG, Gilbert DJ, et al. (2000) Collapsin response mediator protein-3/unc-33-like protein-4 gene: organization, chromosomal mapping and expression in the developing mouse brain. Gene 242: 175–182.
  72. 72. Rosenmund C, Sigler A, Augustin I, Reim K, Brose N, et al. (2002) Differential control of vesicle priming and short-term plasticity by Munc13 isoforms. Neuron 33: 411–424.
  73. 73. Sánchez-Camacho C, Bovolenta P (2008) Autonomous and non-autonomous Shh signalling mediate the in vivo growth and guidance of mouse retinal ganglion cell axons. Development 135: 3531–3541.
  74. 74. Sharma AK, Rohrer B (2007) Sustained elevation of intracellular cGMP causes oxidative stress triggering calpain-mediated apoptosis in photoreceptor degeneration. Current Eye Research 32: 259–269.
  75. 75. Shaw RM, Fay AJ, Puthenveedu MA, von Zastrow M, Jan YN, et al. (2007) Microtubule plus-end-tracking proteins target gap junctions directly from the cell interior to adherens junctions. Cell 128: 547–560.
  76. 76. Shichiri M, Ishimaru S, Ota T, Nishikawa T, Isogai T, et al. (2003) Salusins: newly identified bioactive peptides with hemodynamic and mitogenic activities. Nature Medicine 9: 1166–1172.
  77. 77. Shkumatava A, Neumann CJ (2005) Shh directs cell-cycle exit by activating p57Kip2 in the zebrafish retina. EMBO Reports 6: 563–569.
  78. 78. Suh LH, Oster SF, Soehrman SS, Grenningloh G, Sretavan DW (2004) L1/Laminin modulation of growth cone response to EphB triggers growth pauses and regulates the microtubule destabilizing protein SCG10. Journal of Neuroscience 24: 1976–1986.
  79. 79. Sweeney KJ, Clark GD, Prokscha A, Dobyns WB, Eichele G (2000) Lissencephaly associated mutations suggest a requirement for the PAFAH1B heterotrimeric complex in brain development. Mechanisms of Development 92: 263–271.
  80. 80. Tang YP, Wade J (2006) Sexually dimorphic expression of the genes encoding ribosomal proteins L17 and L37 in the song control nuclei of juvenile zebra finches. Brain Research 1126: 102–108.
  81. 81. Thies E, Davenport RW (2003) Independent roles of Rho-GTPases in growth cone and axonal behavior. Journal of Neurobiology 54: 358–369.
  82. 82. Trimarchi JM, Stadler MB, Roska B, Billings N, Sun B, et al. (2007) Molecular heterogeneity of developing retinal ganglion and amacrine cells revealed through single cell gene expression profiling. Journal of Comparative Neurology 502: 1047–1065.
  83. 83. Tsim TY, Wong EY, Leung MS, Wong CC (2004) Expression of axon guidance molecules and their related genes during development and sexual differentiation of the olfactory bulb in rats. Neuroscience 123: 951–965.
  84. 84. Varoqueaux F, Sigler A, Rhee JS, Brose N, Enk C, et al. (2002) Total arrest of spontaneous and evoked synaptic transmission but normal synaptogenesis in the absence of Munc13-mediated vesicle priming. Proceedings of the National Academy of Science USA 99: 9037–9042.
  85. 85. Wilson MJ, Salata MW, Susalka SJ, Pfister KK (2001) Light chains of mammalian cytoplasmic dynein: identification and characterization of a family of LC8 light chains. Cell Motil Cytoskeleton 49: 229–240.
  86. 86. Young-Pearse TL, Bai J, Chang R, Zheng JB, LoTurco JJ, et al. (2007) A critical function for beta-amyloid precursor protein in neuronal migration revealed by in utero RNA interference. Journal of Neuroscience 27: 14459–14469.
  87. 87. Ready DF, Hanson TE, Benzer S (1976) Development of the Drosophila retina, a neurocrystalline lattice. Developmental Biology 53: 217–240.
  88. 88. Liu YH, Jakobsen JS, Valentin G, Amarantos I, Gilmour DT, et al. (2009) A systematic analysis of Tinman function reveals Eya and JAK-STAT signaling as essential regulators of muscle development. Developmental Cell 16: 280–291.
  89. 89. Hamada H, Meno C, Watanabe D, Saijoh Y (2002) Establishment of vertebrate left-right asymmetry. Nature Reviews Genetics 3: 103–113.
  90. 90. Bell R, Hubbard A, Chettier R, Chen D, Miller JP, et al. (2009) A human protein interaction network shows conservation of aging processes between human and invertebrate species. PLos Genetics 5: e1000414.
  91. 91. Sinclair A, Smith C, Western P, McClive P (2002) A comparative analysis of vertebrate sex determination. In: Chadwick D, Goode J, editors. The Genetics and Biology of Sex Determination, Novartis Foundation Symposium. Chichester: John Wiley & Sons. pp. 102–114.
  92. 92. Kielbasa SM, Vingron M (2008) Transcriptional autoregulatory loops are highly conserved in vertebrate evolution. PLoS ONE 3: e3210.
  93. 93. FlyBase http://flybase.org/.
  94. 94. Informatics MG http://www.informatics.jax.org.
  95. 95. Li H, Coghlan A, Ruan J, Coin LJ, Hériché JK, et al. (2006) TreeFam: a curated database of phylogenetic trees of animal gene families. Nucleic Acids Research 34(Database issue): D572–580.
  96. 96. Ruan J, Li H, Chen Z, Coghlan A, Coin LJM, et al. (2008) TreeFam: 2008 Update. Nucleic Acids Research 36(Database issue): D735–740.
  97. 97. HomoloGene http://www.ncbi.nlm.nih.gov/homologene.
  98. 98. Manfroid I, Caubit X, Kerridge S, Fasano L (2004) Three putative murine Teashirt orthologues specify trunk structures in Drosophila in the same way as the Drosophila teashirt gene. Development 131: 1065–1073.
  99. 99. van Heyningen V, Williamson K (2002) PAX6 in sensory development. Human Molecular Genetics 11: 1161–1167.
  100. 100. Brown NL, Patel S, Brzezinski J, Glaser T (2001) Math5 is required for retinal ganglion cell and optic nerve formation. Development 128: 2497–2508.
  101. 101. Wang SW, Kim BS, Ding K, Wang H, Sun D, et al. (2001) Requirement for Math5 in the development of retinal ganglion cells. Genes & Development 15: 24–29.
  102. 102. Seimiya M, Gehring WJ (2000) The Drosophila homeobox gene optix is capable of inducing ectopic eyes by an eyeless-independent mechanism. Development 127: 1879–1886.
  103. 103. Yona G, Dirks W, Rahman S, Lin DM (2006) Effective similarity measures for expression profiles. Bioinformatics 22: 1616–1622.
  104. 104. Griffith OL, Pleasance ED, Fulton DL, Oveisi M, Ester M, et al. (2005) Assessment and integration of publicly available SAGE, cDNA microarray, and oligonucleotide microarray expression data for global coexpression. Genomics 86: 476–488.
  105. 105. Gunsalus KC, Ge H, Schetter AJ, Goldberg DS, Han JD, et al. (2005) Predictive models of molecular machines involved in Caenorhabditis elegans early embryogenesis. Nature 436: 861–865.
  106. 106. Lee HK, Sajdak J, Qin J, Pavlidis P (2004) Coexpression analysis of human genes across many microarray data sets. Genome Research 14: 1085–1094.
  107. 107. PubMed http://www.ncbi.nlm.nih.gov/pubmed/.
  108. 108. Bult C, Eppig JT, Richardson JE, Blake JA, Group atmotMGD (2008) The Mouse Genome Database (MGD): mouse biology and model systems. Nucleic Acids Research 36: D724–728.