Mouse Phenome Database (MPD)

Abstract

The Mouse Phenome Project was launched a decade ago to complement mouse genome sequencing efforts by promoting new phenotyping initiatives under standardized conditions and collecting the data in a central public database, the Mouse Phenome Database (MPD; http://phenome.jax.org). MPD houses a wealth of strain characteristics data to facilitate the use of the laboratory mouse in translational research for human health and disease, helping alleviate problems involving experimentation in humans that cannot be done practically or ethically. Data sets are voluntarily contributed by researchers from a variety of institutions and settings, or in some cases, retrieved by MPD staff from public sources. MPD maintains a growing collection of standardized reference data that assists investigators in selecting mouse strains for research applications; houses treatment/control data for drug studies and other interventions; offers a standardized platform for discovering genotype-phenotype relationships; and provides tools for hypothesis testing. MPD improvements and updates since our last NAR report are presented, including the addition of new tools and features to facilitate navigation and data mining as well as the acquisition of new data (phenotypic, genotypic and gene expression).

Figure 1.

Phenotype data: new layout, viewing options and tools. An MPD summary plot is shown from an aging study measuring IGF-1 levels in males and females of 33 inbred strains at 6 months of age [Yuan1 (28); see also accompanying paper (29)]. From this plot, users have several choices for viewing (see plot options), and there are one-click options for deploying relevant tools specific for this measurement (boxed selections just above the plot), which include finding other phenotype measurements that correlate significantly, finding genes with correlated expression patterns and performing EMMA-corrected genome-wide association analysis (GWAS). For most MPD measurements with at least 20 inbred strains tested, the GWAS option is a direct link to pre-computed results hosted on the UCLA EMMA website (see text). About the plot: strain means are shown (±1 SEM); dotted horizontal lines indicate 1 SD for the overall strain mean (red = female; blue = male). All words in blue are links to tools, other views or additional information. For a continuation of this example, see Figure 2 (GWAS) and Figure 3 (gene expression).

Figure 2.

New GWAS tools. Continuing with the IGF-1 example [Yuan1 (28)], we performed a genome-wide association analysis by clicking on the ‘GWAS’ option above the plot shown in Figure 1 (the relevant part of this panel is shown here for convenience). We found significant EMMA-corrected peaks for females on chromosomes 2, 10, 13, 14, 15, 17 and 18 (lower panel; green stars were manually overlaid on this plot to ensure visibility of peaks). Interestingly Leduc et al. (29) (the data submitter’s publication for Yuan1) found three of the same in silico QTLs using another method, haplotype association mapping (HAM), which employed the Hidden Markov Model and a set of SNPs at 70K locations (includes imputed calls). We chose to concentrate on the most significant peak identified by these methods on Chr 10 in the vicinity of the Igf1 gene. Using other supporting data sets and bioinformatics tools, Leduc et al. narrowed their results from over 45 genes annotated in the QTL interval to 21 candidate genes. We wondered if we might narrow these results even further by using another new MPD tool for finding significant gene expression correlations (see Figure 3). The new GWAS tool is provided in collaboration with the UCLA Computer Science and Bioinformatics group (ZarLab) (30). MPD is not responsible for the UCLA website or EMMA Server (mouse.cs.ucla.edu/emmaserver).

Figure 3.

New gene expression interface. Continuing with the IGF-1 example [Yuan1 (28)] from Figures 1 and 2, we interrogated the Chr 10 QTL interval (84.5–90.0 Mbp) identified by both association mapping methods (HAM and EMMA) using the new MPD gene expression interface. Users can simply click on the correlated probes/genes option above the plot shown in Figure 1 (the relevant part of this panel is shown in the upper-left panel for convenience), or users may choose to begin in the Correlations Center (S14). Of those 21 candidate genes identified by Leduc et al. (29) (see Figure 2), we went on to confirm that 6 of those genes had compelling expression profiles consistent with IGF-1 levels (P < 0.05). Three of these genes are listed in the middle-left panel (see Note 1 for more information about this panel). As indicated, three probesets from two gene expression projects [GX-Su1 (31,32), GX-Tabakoff1 (33)] were identified using various filtering and sorting options that are available with this tool. Users can adjust stringency to broaden or narrow results by selecting P-value cutoff and setting the data point variance (DPV) quality filter where stricter settings will omit probesets that have larger standard errors relative to the range of strain means. Users may further filter results by minimum number of strains, sex, specific data set and chromosome. Results can be sorted by P-value, correlation coefficient, location order, MPD project or MPD category. A phenotype-gene expression scatterplot is available through the ‘Plot’ link. IGF-1 versus Nr1h4 for males is shown in the lower-left panel. By clicking on gene name, we accessed the gene directory page (lower-right panel, and see Note 2 below), to deploy the 550K SNP panel [CGD2 (3,6)] specifically for the gene of interest to compare the segregation of phenotype with polymorphisms. Using MPD SNP tools, we confirmed that three of the six genes have polymorphic intronic or UTR regions, consistent with the notion that these genes are differentially regulated. The corresponding SNP results for Nr1h4 are shown (middle-right panel) where strains are grouped according to ‘hi’ (left) or ‘lo’ (right) IGF-1 levels. These strains fall into two major haplotypes (wild-derived strains fall into a third haplotype, not shown). For ease of comparison, an optional view of the male IGF-1 data is shown in the upper-right panel [color-guides from the SNP results to plot: blue = high IGF-1 levels, yellow = low IGF-1 levels (see also, strain distribution in Figure 1)]. All words in blue are links to tools, other views or additional information. While we demonstrate that MPD can help identify (or further verify) candidate genes exhibiting differential expression, we also note that caution should be used in making claims of causality based solely on microarray data, SNPs and functional relevance of positional candidates. All MPD phenotype-gene expression pair-wise correlations are available for bulk download (Download Center: S16). Note 1: The ‘collect gene’ option in the middle of the gene expression results (middle-left panel) is one place where users can take advantage of our new ‘workbook’ feature for paring down and gathering candidate genes. Lists are provided a new window, and checkboxes are supplied to exclude (or include) genes in an updated list. At any time, users may opt to generate a separate clean list of gene symbols or various accession ID options to copy/paste into batch query forms of other bioinformatics resources. Note 2: The gene directory box shown in the bottom–right panel is a hub for information and linkouts to other resources (additional features are added as needed). We provide one-click deployment to SNP tools, finding MPD phenotype measurements that correlate, and information about available gene expression probesets, including tools to compare probesets between gene expression projects and to compare female and male for a single probeset within a project. Users will arrive at a gene directory page from any gene link on the MPD website, from gene searches and from general searches that turn up genes. For Developers wanting to link to our gene directory page, use this URL template: http://phenome.jax.org/db/qp?rtn=markers/details&reqsym=Abcd1, where Abcd1 is gene symbol (For Developers: S22).