Ensemble of Gene Set Enrichment Analyses Function


This is the main function to carry out gene set enrichment analysis using the EGSEA algorithm. This function is aimed to extend the limma-voom pipeline of RNA-seq analysis.


egsea(voom.results, contrasts, logFC = NULL, gs.annots, symbolsMap = NULL,
  baseGSEAs = egsea.base(), minSize = 2, display.top = 20,
  combineMethod = "fisher", combineWeights = NULL, sort.by = "p.adj",
  egsea.dir = NULL, kegg.dir = NULL, logFC.cutoff = 0,
  sum.plot.axis = "p.adj", sum.plot.cutoff = NULL, vote.bin.width = 5,
  num.threads = 4, report = TRUE, print.base = FALSE, verbose = FALSE,
  keep.limma = FALSE, keep.set.scores = FALSE)



list, an EList object generated using the voom function. Entrez Gene IDs should be used as row names.


double, an N x L matrix indicates the contrast of the linear model coefficients for which the test is required. N is number of experimental conditions and L is number of contrasts.


double, an K x L matrix indicates the log2 fold change of each gene for each contrast. K is the number of genes included in the analysis. If logFC=NULL, the logFC values are estimated using the ebayes for each contrast.


list, list of objects of class GSCollectionIndex. It is generated using one of these functions: buildIdx, buildMSigDBIdx, buildKEGGIdx, buildGeneSetDBIdx, and buildCustomIdx.


dataframe, an K x 2 matrix stores the gene symbol of each Entrez Gene ID. It is used for the heatmap visualization. The order of rows should match that of the voom.results. Default symbolsMap=NULL.


character, a vector of the gene set tests that should be included in the ensemble. Type egsea.base to see the supported GSE methods. By default, all supported methods are used.


integer, the minimum size of a gene set to be included in the analysis. Default minSize= 2.


integer, the number of top gene sets to be displayed in the EGSEA report. You can always access the list of all tested gene sets using the returned gsa list. Default is 20.


character, determines how to combine p-values from different GSEA method. Type egsea.combine() to see supported methods.


double, a vector determines how different GSEA methods will be weighted. Its values should range between 0 and 1. This option is not supported currently.


character, determines how to order the analysis results in the stats table. Type egsea.sort() to see all available options.


character, directory into which the analysis results are written out.


character, the directory of KEGG pathway data file (.xml) and image file (.png). Default kegg.dir=paste0(egsea.dir, "/kegg-dir/").


numeric, cut-off threshold of logFC and is used for Sginificance Score and Regulation Direction Calculations. Default logFC.cutoff=0.


character, the x-axis of the summary plot. All the values accepted by the sort.by parameter can be used. Default sum.plot.axis="p.value".


numeric, cut-off threshold to filter the gene sets of the summary plots based on the values of the sum.plot.axis. Default sum.plot.cutoff=NULL.


numeric, the bin width of the vote ranking. Default vote.bin.width=5.


numeric, number of CPU threads to be used. Default num.threads=2.


logical, whether to generate the EGSEA interactive report. It takes longer time to run. Default is True.


logical, whether to write out the results of the individual GSE methods. Default FALSE.


logical, whether to print out progress messages and warnings.


logical, whether to return the results of the limma analysis.


logical, whether to calculate the gene set enrichment scores per sample for the methods that support this option, i.e., "ssgsea".


EGSEA, an acronym for Ensemble of Gene Set Enrichment Analyses, utilizes the analysis results of eleven prominent GSE algorithms from the literature to calculate collective significance scores for gene sets. These methods include: ora, globaltest, plage, safe, zscore, gage, ssgsea, roast, fry, padog, camera and gsva. The ora, gage, camera and gsva methods depend on a competitive null hypothesis while the remaining seven methods are based on a self-contained hypothesis. Conveniently, the algorithm proposed here is not limited to these twelve GSE methods and new GSE tests can be easily integrated into the framework. This function takes the voom object and the contrast matrix as parameters. The results of EGSEA can be seen using the topSets function.

EGSEA report is an interactive HTML report that is generated if report=TRUE to enable a swift navigation through the results of an EGSEA analysis. The following pages are generated for each gene set collection and contrast/comparison:
1. Stats Table page shows the detailed statistics of the EGSEA analysis for the display.top gene sets. It shows the EGSEA scores, individual rankings and additional annotation for each gene set. Hyperlinks to the source of each gene set can be seen in this table when they are available. The "Direction" column shows the regulation direction of a gene set which is calculated based on the logFC, which is either calculated from the limma differential expression analysis or provided by the user. The logFC.cutoff is applied for this calculation. The calculations of the EGSEA scores can be seen in the references section. The method topSets can be used to generate custom Stats Table.
2. Heatmaps page shows the heatmaps of the gene fold changes for the gene sets that are presented in the Stats Table page. Red indicates up-regulation while blue indicates down-regulation. Only genes that appear in the input expression/count matrix are visualized in the heat map. Gene names are coloured based on their statistical significance in the limma differential expression analysis. The "Interpret Results" link below each heat map allows the user to download the original heat map values along with additional statistics from limma DE analysis ( if available) so that they can be used to perform further analysis in R, e.g., customizing the heat map visualization. Additional heat maps can be generated and customized using the method plotHeatmap.
3. Summary Plots page shows the methods ranking plot along with the summary plots of EGSEA analysis. The method plot uses multidimensional scaling (MDS) to visualize the ranking of individual methods on a given gene set collection. The summary plots are bubble plots that visualize the distribution of gene sets based on the EGSEA Significance Score and another EGSEA score (default, p-value). Two summary plots are generated: ranking and directional plots. Each gene set is reprersented with a bubble which is coloured based on the EGSEA ranking (in ranking plots ) or gene set regulation direction (in directional plots) and sized based on the gene set cardinality (in ranking plots) or EGSEA Significance score (in directional plots). Since the EGSEA "Significance Score" is proportional to the p-value and the absolute fold changes, it could be useful to highlight gene sets that have high Significance scores. The blue labels on the summary plot indicate gene sets that do not appear in the top 10 list of gene sets based on the "sort.by" argument (black labels) yet they appear in the top 5 list of gene sets based on the EGSEA "Significance Score". If two contrasts are provided, the rank is calculated based on the "comparison" analysis results and the "Significance Score" is calculated as the mean. If sort.by = NULL, the slot sort.by of the object is used to order gene sets. The method plotSummary can be used to customize the Summary plots by changing the x-axis score and filtering bubbles based on the values of the x-axis. The method plotMethods can be used to generate Methods plots.
4. Pathways page shows the KEGG pathways for the gene sets that are presented in the Stats Table of a KEGG gene set collection. The gene fold changes are overlaid on the pathway maps and coloured based on the gene regulation direction: blue for down-regulation and red for up-regulation. The method plotPathway can be used to generate additional pathway maps. Note that this page only appears if a KEGG gene set collection is used in the EGSEA analysis.
5. Go Graphs page shows the Gene Ontology graphs for top 5 GO terms in each of three GO categories: Biological Processes (BP), Molecular Functions (MF), and Cellular Components (CC). Nodes are coloured based on the default sort.by score where red indicates high significance and yellow indicates low significance. The method plotGOGraph can be used to customize GO graphs by changing the default sorting score and the number of significance nodes that can be visualized. It is recommended that a small number of nodes is selected. Note that this page only appears if a Gene Ontology gene set collection is used, i.e., for the c5 collection from MSigDB or the gsdbgo collection from GeneSetDB.

Finally, the "Interpret Results" hyperlink in the EGSEA report allows the user to download the fold changes and limma analysis results and thus improve the interpretation of the results.
Note that the running time of this function significantly increseas when report = TRUE. For example, the analysis in the example section below was conducted on the $203$ signaling and disease KEGG pathways using a MacBook Pro machine that had a 2.8 GHz Intel Core i7 CPU and 16 GB of RAM. The execution time varied between 23.1 seconds (single thread) to 7.9 seconds (16 threads) when the HTML report generation was disabled. The execution time took 145.5 seconds when the report generation was enabled using 16 threads.


A list of elements, each with two/three elements that store the top gene sets and the detailed analysis results for each contrast and the comparative analysis results.


Monther Alhamdoosh, Milica Ng, Nicholas J. Wilson, Julie M. Sheridan, Huy Huynh, Michael J. Wilson and Matthew E. Ritchie. Combining multiple tools outperforms individual methods in gene set enrichment analyses.

See Also

topSets, egsea.base, egsea.sort, buildIdx, buildMSigDBIdx, buildKEGGIdx, buildGeneSetDBIdx, and buildCustomIdx


# Example of egsea
v = il13.data$voom
contrasts = il13.data$contra
gs.annots = buildIdx(entrezIDs=rownames(v$E), species="human", 
         kegg.updated=FALSE, kegg.exclude = c("Metabolism"))
# set report = TRUE to generate the EGSEA interactive report
gsa = egsea(voom.results=v, contrasts=contrasts,  gs.annots=gs.annots, 
         symbolsMap=v$genes, baseGSEAs=egsea.base()[-c(2,5,6,9,12)], 
			display.top = 5, sort.by="avg.rank", 
         num.threads = 2, report = FALSE)
comments powered by Disqus