GOFFA: gene ontology for functional analysis--a FDA gene ontology tool for analysis of genomic and proteomic data

Abstract

Background: Gene Ontology (GO) characterizes and categorizes the functions of genes and their products according to biological processes, molecular functions and cellular components, facilitating interpretation of data from high-throughput genomics and proteomics technologies. The most effective use of GO information is achieved when its rich and hierarchical complexity is retained and the information is distilled to the biological functions that are most germane to the phenomenon being investigated.

Results: Here we present a FDA GO tool named Gene Ontology for Functional Analysis (GOFFA). GOFFA first ranks GO terms in the order of prevalence for a list of selected genes or proteins, and then it allows the user to interactively select GO terms according to their significance and specific biological complexity within the hierarchical structure. GOFFA provides five interactive functions (Tree view, Terms View, Genes View, GO Path and GO TreePrune) to analyze the GO data. Among the five functions, GO Path and GO TreePrune are unique. The GO Path simultaneously displays the ranks that order GOFFA Tree Paths based on statistical analysis. The GO TreePrune provides a visual display of a reduced GO term set based on a user's statistical cut-offs. Therefore, the GOFFA visual display can provide an intuitive depiction of the most likely relevant biological functions.

Conclusion: With GOFFA, the user can dynamically interact with the GO data to interpret gene expression results in the context of biological plausibility, which can lead to new discoveries or identify new hypotheses.

Availability: GOFFA is available through ArrayTrack softwarehttp://edkb.fda.gov/webstart/arraytrack/.

Figure 1

Schematic overview of GOFFA's data flow. GO terms from the Gene Ontology project and gene identifiers from the Entrez Gene databases are combined and linked in the GOFFA database. Lists of genes or proteins from an experiment are analyzed by five functional modules, Tree View, Terms View, Genes View, GO Path and GO TreePrune.

Figure 2

GOFFA interface and Tree Window – The GOFFA interface contains three panels: the left panel (labeled 1) is for queries, the center panel (labeled 2) for tabular and/or graphical displays of and for interaction with the GO information, and the right panel (labeled 3) lists the individual genes associated with the GO information presented in the center panel. The displayed Tree Window in the center panel is the default view of GOFFA, which enables the hierarchical display of the GO terms in a outline-like tree format; p- and E-values as well as the number of genes are also displayed for each GO term. E-values >1 are shown in green and those <1 in red, respectively denoting greater or lesser prevalence, respectively, of the GO term in the inputted gene list rather than in the overall experimental platform. The user can query the tree by GO term, gene name/symbol, p-value, E-value and in combination with functions below the view. The query-match GO terms are highlighted as blue.

Figure 3

Terms and Genes Windows – The Terms Window (A) and Genes Window (B) summarize the findings associated with GO terms and genes respectively in the tabular format along with various statistical parameters (e.g., p- and E-values). Each View contains three tables corresponding to three categories of GO (molecular functions, biological processes and cellular components). The table can be sorted in every column by clicking on the column header. Sorting on multiple columns is also supported (pressing Ctrl key while clicking on the second column header for sorting). Both copy/paste and export functions are available to transfer data to external software.

Figure 4

GO Path – GO Path sorts, by descending statistical significance based on an inverse Chi-Squared test, the GOFFA Tree Paths (i.e., linked GO terms) and graphically displays them from high to low at each hierarchical level. GO Path plots the top ten paths with solid circles representing the GO terms on the path. The X-axis has the hierarchical level to which the GO term belongs and the Y-axis (log p) indicates the statistical significance of the term. A color key for the top 10 paths (as determined by equation 2) is located beneath the plot. Clicking either a circle in a path in the plot or its corresponding color key launches a Tree View (Figure 2) with the selected path highlighted in blue. Other features are also available from a popup menu obtained by right clicking the plot, including zoom in/out, export figure, etc.

Figure 5

GO TreePrune – This node-like tree display allows the user to filter out nodes and thus reduce the complexity of a tree by specifying the p- and E-value as well as the user-defined number of genes in the end node. A GO term is represented by a sectored pie, where the red sector shows the percentage of the inputted genes associated with the term. The individual genes associated with each term are displayed in the right panel by single clicking the term. The annotation of a term can be turned on or off by double clicking the term. Each term is movable with mouse drag, which is convenient when working on a dense tree or with many annotations. The tree diagram can be zoomed and moved by holding down the right or left button of the mouse, respectively.