Contents

Introduction

The main objective of ViSEAGO package is to carry out a data mining of biological functions and establish links between genes involved in the study. We developed ViSEAGO in R to facilitate functional Gene Ontology (GO) analysis of complex experimental design with multiple comparisons of interest. It allows to study large-scale datasets together and visualize GO profiles to capture biological knowledge. The acronym stands for three major concepts of the analysis: Visualization, Semantic similarity and Enrichment Analysis of Gene Ontology. It provides access to the last current GO annotations, which are retrieved from one of NCBI EntrezGene, Ensembl or Uniprot databases for several species. Using available R packages and novel developments, ViSEAGO extends classical functional GO analysis to focus on functional coherence by aggregating closely related biological themes while studying multiple datasets at once. It provides both a synthetic and detailed view using interactive functionalities respecting the GO graph structure and ensuring functional coherence supplied by semantic similarity. ViSEAGO has been successfully applied on several datasets from different species with a variety of biological questions. Results can be easily shared between bioinformaticians and biologists, enhancing reporting capabilities while maintaining reproducibility.

package diagram

ViSEAGO is like a workflow with input lists of genes and associated genomic resources in order to perform enrichment tests, followed by a clustering of GO terms based on theirs semantic similarity. Most of outputs (tables and graphs) could be visualized in static or interactive mode.

workflow

Installation

Firtsly, install the ViSEAGO R package from Bioconductor, or ForgeMIA gitlab.

NB: This vignette is not runnable and provide a pseudo-code to illustrate ViSEAGO package capabilities.

1 Genes of interest

Firstly, you need to provide the gene background (= expressed genes) and selection files.

2 GO annotation of genes

The Gene Ontology (GO) is a community-based bioinformatics resource that supplies information about gene product function using ontologies to represent biological knowledge [1].

Currently, the 3 predominant genomic ressources are EntrezGene [2], Ensembl [3], and Uniprot-GOA [4]. Furthermore, an easy access with the R langage [5] to some genomic GO annotations from EntrezGene is available using Bioconductor [6] org.xx.eg.db databases. However, there are important differences between supported species, and corresponding genes knowledge according to database used. This has necessary an impact on GO annotations and enrichment and downstream analyses.

More precisely, since December 2006, the GOA and Ensembl Compara teams have used the gene orthology data obtained from Ensembl Compara to project GO terms from a source species onto one or more target species. For this pipeline, only one to one and apparent one to one orthologies are used, and only manually annotated GO terms with an experimental evidence type of either EXP, IDA, IEP, IGI, IMP or IPI are projected. GO annotations created using this technique receive the evidence code ‘IEA’ (Inferred from Electronic Annotation).

A similar approach for EntrezGene database and with the orthologs_groups [7] from the NCBI annotation pipeline. This method is available using the ViSEAGO::annotate(id,EntrezGene,ortholog = TRUE).

The easy access to GO annotations for all the availables species from these databases are provided by:

In order to know if your target species is supported by the database, we can display the available organisms in an interactive table using ViSEAGO::available_organisms method. The organism key identifiant is always displayed in the first column of the table.

At this step, we need to provide the organism key identifiant based on the genomic ressource used (Bioconductor, EntrezGene, Ensembl, or Uniprot), in order to extract and store the species GO annotations using ViSEAGO::annotate() method.

3 Functionnal GO enrichment

3.1 GO enrichment tests

The GO terms enrichment tests are performed with the topGO package [8], which support several algorithms that improve GO group scoring using the underlying GO graph topology.

First, we create a topGOdata object, using ViSEAGO::create_topGOdata method, with inputs as: genes selection, genes background, GO terms category used (MF, BP, or CC), and minimum of annotated genes by GO terms (nodeSize).

Now, we perform the GO enrichment tests for BP category with Fisher’s exact test with one of the four available algorithms (classic, elim, weight, and weight01) using topGO::runTest method.

NB: p-values of GO terms are not adjusted and considered significant below 0.01 (see topGO vignette).

3.2 Combine enriched GO terms

We combine results of enrichment tests into an object using ViSEAGO::merge_enrich_terms method. A table of enriched GO terms in at least one comparison is displayed in interactive mode, or printed in a file using ViSEAGO::show_table method. The printed table contains for each enriched GO terms, additional columns including the list of significant genes and frequency (ratio of number of significant genes to number of background genes) evaluated by comparison.

3.3 Graphs of GO enrichment tests

Several graphs summarize important features. An interactive barchart showing the number of GO terms enriched or not in each comparison, using ViSEAGO::GOcount method. Intersections of lists of enriched GO terms between comparisons are displayed in an Upset plot using ViSEAGO::Upset method.

4 GO terms Semantic Similarity

GO annotations of genes created at a previous step and enriched GO terms are combined using ViSEAGO::build_GO_SS method. The Semantic Similarity (SS) between enriched GO terms are calculated using ViSEAGO::compute_SS_distances method which is a wrapper of functions implemented in the GOSemSim package [9]. Among the SS algorithms between GO terms, 4 of them use information content (IC) of a GO term which is computed by the negative log probability of the term occurring in GO corpus. A rarely used term contains a greater amount of information. These IC-based mehods are Resnik, Lin, Rel, and Jiang. A Graph-based method, using the topology of GO graph structure to compute semantic similarity is also available with the Wang method. The built object myGOs contains all informations of enriched GO terms and computed SS distances between them.

5 Visualization and interpretation of enriched GO terms

5.1 Multi Dimensional Scaling of GO terms - A preview

A Multi Dimensional Scale (MDS) plot with ViSEAGO::MDSplot method provides a representation of distances among a set of enriched GO terms on the two first dimensions. Some patterns could appear at this time and could be investigated in an interactive mode. The plot could be also printed in a png file.

5.2 Clustering heatmap of GO terms

To fully explore the results of this functional analysis, a hierarchical clustering method using ViSEAGO::GOterms_heatmap is performed based on one of SS distances (i.e Wang) between the enriched GO terms and a chosen aggregation criteria (i.e ward.D2).

Clusters of enriched GO terms are obtained by cutting branches off the dendrogram. Here, we choose a dynamic branch cutting method based on the shape of clusters using dynamicTreeCut [10,11]. It builds clusters in a bottom top manner in two steps. In step 1, branches that satisfy specific criteria for being clusters, based on the merging information of the dendrogram, are detected. A simple control of the relative sensitivity to cluster splitting is provided in the parameter deepSplit (The higher the value, the more and smaller clusters will be produced), and the minimum cluster size with minClusterSize. In step 2, all previously unassigned objects are tested for sufficient proximity to clusters detected in the first step using a modified Partitioning Around Medoids (PAM). If the closest cluster is close enough in a precise sense that we define below, the object is assigned to that cluster. The PAM stage will be performed using pamStage=TRUE parameter, and respect the dendrogram using pamRespectsDendro=TRUE parameter, in the sense an object can be PAM-assigned only to clusters that lie below it on the branch that the object is merged into.

Enriched GO terms are ranked in a dendrogram as a result of a hierarchical clustering and colored depending on their cluster assignation. Additional illustrations are displayed: the GO description of GO terms (trimmed if more than 50 characters), a heatmap of -log10(pvalue) of enrichment test for each comparison, and optionally the Information Content (IC). The illustration is displayed with ViSEAGO::show_heatmap method.

A table of enriched GO terms in at least one of the comparison is displayed in interactive mode, or printed in a file using ViSEAGO::show_table method. The columns GO cluster number and IC value are added to the previous table of enriched terms table. The printed table contains for each enriched GO terms, additional columns including the list of significant genes and frequency (ratio between the number of significant genes and number of background genes in a specific GO tag) evaluated by comparison.

5.3 Multi Dimensional Scaling of GO terms

We also display a Multi Dimensional Scale (MDS) plot with the overlay of GO terms clusters from Wang_clusters_wardD2 object with ViSEAGO::MDSplot method. It is a way to check the coherence of GO terms clusters on the MDS plot.

6 Visualization and interpretation of GO clusters

This part is dedicated to explore relationships between clusters of GO terms defined by the last hierarchical clustering with ViSEAGO::GOterms_heatmap method. This analysis should be conducted if the number of clusters of GO terms is large enough and therefore difficult to investigate.

6.1 Compute semantic similiarity between GO clusters

A Semantic Similarity (SS) between clusters of GO terms, previously defined, are calculated using ViSEAGO::compute_SS_distances method. The availables SS algorithms between GO terms clusters, implemented in the GOSemSim package [9], could be max, average, rcmax, or BMA.

A colored Multi Dimensional Scale (MDS) plot with ViSEAGO::MDSplot method provides a representation of distances between the clusters of GO terms. Each circle represents a cluster of GO terms and its size depends on the number of GO terms that it contains. Clusters of GO terms that are close should share a functional coherence.

6.2 GO clusters semantic similarities heatmap

A new hierarchical clustering using ViSEAGO::GOclusters_heatmap is performed based on one of the SS distance (i.e. BMA) between the clusters of GO terms, previously computed, and a chosen aggregation criteria (i.e ward.D2).

Clusters of GO terms are ranked in a dendrogram and branches are colored depending on the cluster assignation. Additional illustrations are displayed: definition of the cluster of GO terms corresponds to the first common GO term ancestor followed by the cluster label in brackets, and heatmap of GO terms count from the corresponding cluster. The illustration is displayed with ViSEAGO::show_heatmap method.

7 Conclusion

A complete functional Gene Ontology (GO) enrichment analysis can be performed using ViSEAGO package. It facilitates the interpretation of biological processes, functions, or cellular component involved in the study. It provides both a synthetic and detailed view using interactive functionalities respecting GO graph structure and ensuring functional coherence. Also, comparisons of enrichment can be made on large datasets and for complex experimental designs, such as explained in mouse mammary gland status comparison example (see utils::vignette("2_mouse_bionconductor",package ="ViSEAGO")). Moreover, we can explore the effect of the GO semantic similarity algorithms on the tree structure, such as illustrate in the clustering comparions example (seevignette("3_SS_choice",package ="ViSEAGO")).

Information session

               _                           
platform       x86_64-pc-linux-gnu         
arch           x86_64                      
os             linux-gnu                   
system         x86_64, linux-gnu           
status                                     
major          3                           
minor          6.1                         
year           2019                        
month          07                          
day            05                          
svn rev        76782                       
language       R                           
version.string R version 3.6.1 (2019-07-05)
nickname       Action of the Toes          

References

1. Blake JA, Christie KR, Dolan ME, Drabkin HJ, Hill DP, Ni L, et al. Gene Ontology Consortium: going forward. Nucleic Acids Res. 2015;43:D1049–1056.

2. Brown GR, Hem V, Katz KS, Ovetsky M, Wallin C, Ermolaeva O, et al. Gene: a gene-centered information resource at NCBI. Nucleic Acids Res. 2015;43:36–42.

3. Aken BL, Achuthan P, Akanni W, Amode MR, Bernsdorff F, Bhai J, et al. Ensembl 2017. Nucleic Acids Res. 2017;45:D635–42.

4. Huntley RP, Sawford T, Mutowo-Meullenet P, Shypitsyna A, Bonilla C, Martin MJ, et al. The GOA database: gene Ontology annotation updates for 2015. Nucleic Acids Res. 2015;43:D1057–1063.

5. R Core Team. R: A language and environment for statistical computing [Internet]. Vienna, Austria: R Foundation for Statistical Computing; 2017. Available from: https://www.R-project.org/

6. Huber W, Carey VJ, Gentleman R, Anders S, Carlson M, Carvalho BS, et al. Orchestrating high-throughput genomic analysis with Bioconductor. Nat Methods. 2015;12:115–21.

7. Fong JH, Murphy TD, Pruitt KD. Comparison of RefSeq protein-coding regions in human and vertebrate genomes. BMC Genomics. 2013;14:654.

8. Alexa A, Rahnenfuhrer J, Lengauer T. Improved scoring of functional groups from gene expression data by decorrelating GO graph structure. Bioinformatics. 2006;22:1600–7.

9. Yu G, Li F, Qin Y, Bo X, Wu Y, Wang S. GOSemSim: an R package for measuring semantic similarity among GO terms and gene products. Bioinformatics. 2010;26:976–8.

10. Langfelder P, Zhang B, Horvath S. Defining clusters from a hierarchical cluster tree: the Dynamic Tree Cut package for R. Bioinformatics. 2008;24:719–20.

11. Langfelder P, Zhang B, Steve Horvath. DynamicTreeCut: Methods for detection of clusters in hierarchical clustering dendrograms [Internet]. 2016. Available from: https://CRAN.R-project.org/package=dynamicTreeCut