Figure 1.
GenAMap is a visual analytics system for structured association mapping.
Through the UI, users can explore the population and network structures of the data and determine which association analyses to run. Users can also take advantage of new, intuitive visualizations to explore the structures inherent in the data while simultaneously exploring the results from association analysis. All jobs are run on a remote cluster, and the results are displayed from the front end, linking out to external databases for further information and analysis.
Figure 2.
GenAMap is run locally as a desktop application. It communicates directly with our cluster through Auto-SAM, an automatic system for running structured association algorithms. GenAMap executes all tasks, returning to the user a set of visualizations to explore and analyze the results to find interesting signals in the data.
Table 1.
Algorithms available to run in GenAMap.
Table 2.
Major data set types available to import/create via an algorithm in GenAMap.
Figure 3.
GenAMap provides a simple genome browser that allows analysts to explore the mutation marker data that they load into GenAMap. SNPs are represented by green circles across the genome. Analysts can use these SNPs to directly link to external databases, such as SGD or dbSNP. SNP labels are displayed as the analyst hovers over the SNPs.
Figure 4.
GenAMap trait overview exploration.
GenAMap provides an overview of gene and phenotypic trait networks to aid analysts in their exploration of the networks. Here, we present a genetic network generated from the yeast data. The network has been further organized by hierarchical clustering, and twenty highly connected gene modules have been automatically identified by GenAMap (outlined in color). As the analyst clicks in these different modules, an information display appears to report the GO and eQTL enrichment of the genes that belong to the particular module.
Figure 5.
Using GenAMap to explore genetic networks.
We demonstrate using GenAMap visualizations to explore a genetic network. A) From the overview of the network, the analyst can see the different gene modules in the network. B) The analyst zooms into a module of interest in the network. C) The analyst switches to a node-edge representation of this sub-network and adjusts the edge threshold, layout, and labels. D) The analyst uses GenAMap to link directly to external data sources for more information.
Figure 6.
GenAMap overview of association results.
GenAMap provides a heat chart visualization to explore the results from an eQTL association analysis. SNPs are plotted along the x-axis and genes are clustered along the y-axis. This view allows the analyst to explore the overview of the results. For example, in these results from running TreeLasso on the yeast data, many SNPs are associated with all the genes in a gene module, and some gene modules are associated with many different SNPs in different genomic locations.
Figure 7.
Using GenAMap to find eQTLs in yeast data.
GenAMap provides many tools for analysts to explore association results while using the structure of the data to guide the discovery of associations. We demonstrate some of these tools. A) The analyst can zoom into certain regions to see finer detail of the SNP-phenotypic trait associations. This panel is a zoomed-in region from Figure 6. B) The analyst switches to the JUNG view to explore the genes associated with the region and perform a GO enrichment test. C) The analyst colors the genes by strength of association to the genomic region. D) The analyst selects up to ten interesting genes (salmon colored) and views the Manhattan plot of associations from these genes across the genome. E) The analyst zooms into interesting regions in the genome view. F) The analyst can switch between association tests for further insight into the associations.
Table 3.
Comparison of gene networks across mouse tissues.
Table 4.
Gene modules with GO enrichment in the liver network.
Figure 8.
Association of axon genes to chromosome 14.
We found that rs8244120 on chromosome 14 was associated with 140 genes enriched for cell projection, implying function in neuronal axons. Here, we show 22 of the genes with the strongest associations in GenAMap’s node-link view, colored by the strength of association to rs8244120. White genes are strongly associated and black genes are weakly associated (gray is intermediate). We found that some of the genes were also associated with another SNP on chromosome 14 (shown) and some of the genes were associated with a SNP on chromosome 18 (not shown).
Table 5.
GFlasso-gGFlasso results for the association analysis of the mice dataset.
Figure 9.
Overview of three way GFlasso-gGFlasso association analysis.
We show the overview of the phenotypic trait-network and gene-network from GenAMap for the GFlasso-gGFlasso analysis; associations are not shown. In this visualization, circles represent groups of genes, associated to the same regions in the genome. Hexagons represent phenotypic traits. The edges between genes or between phenotypic traits represent the connections in the gene or trait network. In this data, we note that there are very few edges between gene groups. The largest gene group is the teal group, representing genes associated with the eQTL hotspot on chromosome 14. The phenotypic trait network consists of small sub-groups of related traits.
Figure 10.
Analyzing population structure in GenAMap.
GenAMap provides an interactive view for analysts to explore population structure. Population assignments are plotted by individual by Eigenvalue. The analyst can adjust the 2D plot to adjust between the first five Eigenvalues. Here, we present the results from a population analysis on the mouse data. The population label for each individual predicted by Structure is plotted according to the first two Eigenvalues. The plot shows clear separation between the populations.
Figure 11.
Interactive Manhattan plot for population data.
GenAMap provides an interactive Manhattan plot for exploring associations in population data. Here we show the results of MPGL looking for genetic associations associated with the gene expression Mapk1 in the hippocampus gene expression data. The four colors represent the four populations in the data detected by Structure. The strength of the association in population 1 is significantly higher for all SNPs than in the other 3 populations.
Figure 12.
Frequency distribution of phenotypic trait by genotype.
When exploring SNP-gene expression associations, GenAMap provides links to tools that allow the analyst to explore the discovered association. For example, consider a case where the analyst considers a discovered SNP-phenotypic trait association. The analyst can query dbSNP to find out information about the SNP, and the analyst can use GenAMap to visualize the frequency distribution of the phenotypic trait by genotype.