Seamless navigation through combined results of set- & network-based enrichment analysis
06 May, 2024
Abstract
The package EnrichmentBrowser implements an analysis pipeline for high-throughput gene expression data as measured with microarrays and RNA-seq. In a workflow-like manner, the package brings together a selection of established Bioconductor packages for gene expression data analysis. It integrates a wide range of gene set and network enrichment analysis methods and facilitates combination and exploration of results across methods.
Package
EnrichmentBrowser 2.34.1
Report issues on https://github.com/lgeistlinger/EnrichmentBrowser/issues
1 Introduction
The package EnrichmentBrowser implements essential functionality for the enrichment analysis of gene expression data. The analysis combines the advantages of set-based and network-based enrichment analysis to derive high-confidence gene sets and biological pathways that are differentially regulated in the expression data under investigation. Besides, the package facilitates the visualization and exploration of such sets and pathways. The following instructions will guide you through an end-to-end expression data analysis workflow including:
Preparing the data
Preprocessing of the data
Differential expression (DE) analysis
Defining gene sets of interest
Executing individual enrichment methods
Combining the results of different methods
Visualize and explore the results
All of these steps are modular, i.e. each step can be executed individually and fine-tuned with several parameters. In case you are interested in a particular step, you can directly move on to the respective section. For example, if you have differential expression already calculated for each gene, and you are now interested in whether certain gene functions are enriched for differential expression, section 7.2 would be the one you should go for. The last section 9 also demonstrates how to wrap the whole workflow into a single function, making use of suitably chosen defaults.
2 Reading expression data from file
Typically, the expression data is not already available in R but
rather has to be read in from a file. This can be done using the function
readSE, which reads the expression data (exprs) along with the
phenotype data (colData) and feature data (rowData) into a
SummarizedExperiment.
library(EnrichmentBrowser)
data.dir <- system.file("extdata", package = "EnrichmentBrowser")
exprs.file <- file.path(data.dir, "exprs.tab")
cdat.file <- file.path(data.dir, "colData.tab")
rdat.file <- file.path(data.dir, "rowData.tab")
se <- readSE(exprs.file, cdat.file, rdat.file)The man pages provide details on file format and the SummarizedExperiment data structure.
?readSE
?SummarizedExperimentNote: Previous versions of the EnrichmentBrowser used the ExpressionSet data structure. The migration to SummarizedExperiment in the current release of the EnrichmentBrowser is done to reflect recent developments in Bioconductor, which discourages the use of ExpressionSet in favor of SummarizedExperiment. The major reasons are the compatibility of SummarizedExperiment with operations on genomic regions as well as efficient handling of big data.
To enable a smooth transition, all functions of the EnrichmentBrowser are still accepting also an ExpressionSet as input, but are consistently returning a SummarizedExperiment as output.
Furthermore, users can always coerce the SummarizedExperiment to ExpressionSet via
eset <- as(se, "ExpressionSet")and vice versa
se <- as(eset, "SummarizedExperiment")3 Types of expression data
The two major data types processed by the EnrichmentBrowser are microarray (intensity measurements) and RNA-seq (read counts) data.
Although RNA-seq has become the de facto standard for transcriptomic profiling, it is important to know that many methods for differential expression and gene set enrichment analysis have been originally developed for microarray data.
However, differences in data distribution assumptions (microarray: quasi-normal, RNA-seq: negative binomial) made adaptations in differential expression analysis and, to some extent, also in gene set enrichment analysis if necessary.
Thus, we consider two example data sets – a microarray and an RNA-seq data set, and discuss similarities and differences between the respective analysis steps.
3.1 Microarray data
To demonstrate the package’s functionality for microarray data, we consider expression measurements of patients with acute lymphoblastic leukemia [1]. A frequent chromosomal defect found among these patients is a translocation, in which parts of chromosome 9 and 22 swap places. This results in the oncogenic fusion gene BCR/ABL created by positioning the ABL1 gene on chromosome 9 to a part of the BCR gene on chromosome 22.
We load the ALL dataset
library(ALL)
data(ALL)and select B-cell ALL patients with and without the BCR/ABL fusion as described previously [2].
ind.bs <- grep("^B", ALL$BT)
ind.mut <- which(ALL$mol.biol %in% c("BCR/ABL", "NEG"))
sset <- intersect(ind.bs, ind.mut)
all.eset <- ALL[, sset]We can now access the expression values, which are intensity measurements on a log-scale for 12,625 probes (rows) across 79 patients (columns).
dim(all.eset)
#> Features Samples
#> 12625 79
exprs(all.eset)[1:4,1:4]
#> 01005 01010 03002 04007
#> 1000_at 7.597323 7.479445 7.567593 7.905312
#> 1001_at 5.046194 4.932537 4.799294 4.844565
#> 1002_f_at 3.900466 4.208155 3.886169 3.416923
#> 1003_s_at 5.903856 6.169024 5.860459 5.687997As we often have more than one probe per gene, we summarize the gene expression values as the average of the corresponding probe values.
allSE <- probe2gene(all.eset)
head(rownames(allSE))
#> [1] "5595" "7075" "1557" "643" "1843" "4319"Note, that the mapping from the probe to gene is done automatically as long
as you have the corresponding annotation package, here the
hgu95av2.db package, installed. Otherwise, the mapping
can be manually defined in the rowData slot.
rowData(se)
#> DataFrame with 1000 rows and 1 column
#> ENTREZID
#> <character>
#> 3075 3075
#> 572 572
#> 4267 4267
#> 26 26
#> 51384 51384
#> ... ...
#> 5295 5295
#> 2966 2966
#> 9140 9140
#> 5558 5558
#> 1956 19563.2 RNA-seq data
To demonstrate the functionality of the package for RNA-seq data, we consider transcriptome profiles of four primary human airway smooth muscle cell lines in two conditions: control and treatment with dexamethasone [3].
We load the airway dataset
library(airway)
data(airway)For further analysis, we remove genes with very low read counts and measurements that are not mapped to an ENSEMBL gene ID.
airSE <- airway[grep("^ENSG", rownames(airway)),]
airSE <- airSE[rowSums(assay(airSE)) > 4,]
dim(airSE)
#> [1] 25133 8
assay(airSE)[1:4,1:4]
#> SRR1039508 SRR1039509 SRR1039512 SRR1039513
#> ENSG00000000003 679 448 873 408
#> ENSG00000000419 467 515 621 365
#> ENSG00000000457 260 211 263 164
#> ENSG00000000460 60 55 40 354 Normalization
Normalization of high-throughput expression data is essential to make
results within and between experiments comparable. Microarray (intensity
measurements) and RNA-seq (read counts) data typically show distinct
features that need to be normalized for. The function normalize wraps
commonly used functionality from limma for microarray
normalization and from EDASeq for RNA-seq normalization.
For specific needs that deviate from these standard normalizations, the
user should always refer to more specific functions/packages.
Microarray data is expected to be single-channel. For two-color arrays,
it is expected that normalization within arrays has been already carried
out, e.g. using from normalizeWithinArrays from limma.
A default quantile normalization based on normalizeBetweenArrays from
limma can be carried out via
allSE <- normalize(allSE, norm.method = "quantile")par(mfrow=c(1,2))
boxplot(assay(allSE, "raw"))
boxplot(assay(allSE, "norm"))
Note that this is only done for demonstration, as the ALL data has been
already RMA-normalized by the authors of the ALL dataset.
RNA-seq data is expected to be raw read counts. Note that normalization for downstream DE analysis, e.g.with edgeR and DESeq2, is not ultimately necessary (and in some cases even discouraged) as many of these tools implement specific normalization approaches themselves. See the vignette of EDASeq, edgeR, and DESeq2 for details.
In case normalization is desired, between-lane normalization to adjust for sequencing depth can be carried out as demonstrated for microarray data.
airSE <- normalize(airSE, norm.method = "quantile")Within-lane normalization to adjust for gene-specific effects such as
gene length and GC-content require to retrieve this information first,
e.g. from BioMart or specific Bioconductor annotation packages. Both
modes are implemented in the EDASeq function
getGeneLengthAndGCContent.
5 Differential expression
The EnrichmentBrowser incorporates established
functionality from the limma package for differential
expression analysis between sample groups. This involves the
voom-transformation when applied to RNA-seq data. Alternatively,
differential expression analysis for RNA-seq data can also be carried
out based on the negative binomial distribution with
edgeR and DESeq2.
This can be performed using the function deAna and assumes some
standardized variable names:
GROUP defines the sample groups being contrasted,
BLOCK defines paired samples or sample blocks, e.g. for batch effects.
For more information on experimental design, see the limma user’s guide, chapter 9.
For the ALL dataset, the GROUP variable indicates whether the BCR-ABL gene fusion is present (1) or not (0).
allSE$GROUP <- ifelse(allSE$mol.biol =="BCR/ABL", 1, 0)
table(allSE$GROUP)
#>
#> 0 1
#> 42 37For the airway dataset, it indicates whether the cell lines have been treated with dexamethasone (1) or not (0).
airSE$GROUP <- ifelse(airway$dex == "trt", 1, 0)
table(airSE$GROUP)
#>
#> 0 1
#> 4 4Paired samples, or in general sample batches/blocks, can be defined via
a BLOCK column in the colData slot. For the airway dataset, the
sample blocks correspond to the four different cell lines.
airSE$BLOCK <- airway$cell
table(airSE$BLOCK)
#>
#> N052611 N061011 N080611 N61311
#> 2 2 2 2For microarray expression data, the deAna function carries out a
differential expression analysis between the two groups based on
functionality from the limma package. Resulting fold
changes and t-test derived p-values for each gene are appended to
the rowData slot.
allSE <- deAna(allSE, padj.method = "BH")
rowData(allSE)
#> DataFrame with 9054 rows and 4 columns
#> FC limma.STAT PVAL ADJ.PVAL
#> <numeric> <numeric> <numeric> <numeric>
#> 5595 0.0391102 0.663788 0.5087392 0.859540
#> 7075 0.0165359 0.234715 0.8150313 0.957808
#> 1557 -0.0502313 -1.267260 0.2087485 0.687276
#> 643 -0.0305410 -0.662652 0.5094628 0.859540
#> 1843 -0.4139776 -1.764693 0.0814421 0.512067
#> ... ... ... ... ...
#> 6300 -0.0450580 -0.946760 0.346619 0.781975
#> 7297 -0.1340087 -1.231891 0.221608 0.700189
#> 2246 0.0309975 0.799858 0.426168 0.819762
#> 7850 -0.0214471 -0.245232 0.806907 0.957327
#> 1593 -0.0126749 -0.254181 0.800009 0.956083Nominal p-values (PVAL) are corrected for multiple testing
(ADJ.PVAL) using the method from Benjamini and Hochberg implemented in
the functionp.adjust from the stats package.
To get a first overview, we inspect the p-value distribution and the volcano plot (fold change against p-value).
par(mfrow = c(1,2))
pdistr(rowData(allSE)$PVAL)
volcano(rowData(allSE)$FC, rowData(allSE)$ADJ.PVAL)
The expression change of the highest statistical significance is observed for the ENTREZ gene 7525.
ind.min <- which.min(rowData(allSE)$ADJ.PVAL)
rowData(allSE)[ind.min,]
#> DataFrame with 1 row and 4 columns
#> FC limma.STAT PVAL ADJ.PVAL
#> <numeric> <numeric> <numeric> <numeric>
#> 7525 1.42179 7.01979 6.5712e-10 5.94956e-06This turns out to be the YES proto-oncogene 1 (hsa:7525@KEGG).
For RNA-seq data, the deAna function carries out a differential
expression analysis between the two groups either based on functionality
from limma (that includes the voom transformation), or
alternatively, the popular edgeR or DESeq2
package.
Here, we use the analysis based on edgeR for demonstration.
airSE <- deAna(airSE, de.method = "edgeR")
#> Excluding 9207 genes not satisfying min.cpm threshold
rowData(airSE)
#> DataFrame with 15926 rows and 14 columns
#> gene_id gene_name entrezid gene_biotype
#> <character> <character> <integer> <character>
#> ENSG00000000003 ENSG00000000003 TSPAN6 NA protein_coding
#> ENSG00000000419 ENSG00000000419 DPM1 NA protein_coding
#> ENSG00000000457 ENSG00000000457 SCYL3 NA protein_coding
#> ENSG00000000460 ENSG00000000460 C1orf112 NA protein_coding
#> ENSG00000000971 ENSG00000000971 CFH NA protein_coding
#> ... ... ... ... ...
#> ENSG00000273373 ENSG00000273373 RP5-1074L1.4 NA antisense
#> ENSG00000273382 ENSG00000273382 RP5-1065J22.8 NA antisense
#> ENSG00000273448 ENSG00000273448 RP11-166O4.6 NA lincRNA
#> ENSG00000273472 ENSG00000273472 RP11-102N12.3 NA lincRNA
#> ENSG00000273486 ENSG00000273486 RP11-731C17.2 NA antisense
#> gene_seq_start gene_seq_end seq_name seq_strand
#> <integer> <integer> <character> <integer>
#> ENSG00000000003 99883667 99894988 X -1
#> ENSG00000000419 49551404 49575092 20 -1
#> ENSG00000000457 169818772 169863408 1 -1
#> ENSG00000000460 169631245 169823221 1 1
#> ENSG00000000971 196621008 196716634 1 1
#> ... ... ... ... ...
#> ENSG00000273373 110912776 110915625 1 1
#> ENSG00000273382 109630593 109633480 1 -1
#> ENSG00000273448 66798034 66799370 7 1
#> ENSG00000273472 141677682 141679075 4 1
#> ENSG00000273486 136556180 136557863 3 -1
#> seq_coord_system symbol FC edgeR.STAT
#> <integer> <character> <numeric> <numeric>
#> ENSG00000000003 NA TSPAN6 -0.390063 31.5879445
#> ENSG00000000419 NA DPM1 0.197780 6.5770644
#> ENSG00000000457 NA SCYL3 0.029173 0.0956747
#> ENSG00000000460 NA C1orf112 -0.137318 0.4832384
#> ENSG00000000971 NA CFH 0.417266 29.6690052
#> ... ... ... ... ...
#> ENSG00000273373 NA RP5-1074L1.4 -0.0498788 0.05448881
#> ENSG00000273382 NA RP5-1065J22.8 -0.9042381 8.79228935
#> ENSG00000273448 NA RP11-166O4.6 0.0197439 0.00531262
#> ENSG00000273472 NA RP11-102N12.3 -0.4753578 2.02231540
#> ENSG00000273486 NA RP11-731C17.2 -0.0959258 0.11916294
#> PVAL ADJ.PVAL
#> <numeric> <numeric>
#> ENSG00000000003 0.000271743 0.00251469
#> ENSG00000000419 0.029335639 0.07942866
#> ENSG00000000457 0.763789843 0.84437853
#> ENSG00000000460 0.503681078 0.63785185
#> ENSG00000000971 0.000342796 0.00299801
#> ... ... ...
#> ENSG00000273373 0.8204049 0.8821070
#> ENSG00000273382 0.0150524 0.0484195
#> ENSG00000273448 0.9434164 0.9654212
#> ENSG00000273472 0.1872562 0.3156813
#> ENSG00000273486 0.7375056 0.82598556 ID mapping
Using genomic information from different resources often requires mapping between different types of gene identifiers. Although primary analysis steps such as normalization and differential expression analysis can be carried out independent of the gene ID type, downstream exploration functionality of the EnrichmentBrowser is consistently based on NCBI Entrez Gene IDs. It is thus, in this regard, beneficial to initially map gene IDs of a different type to NCBI Entrez IDs.
The function idTypes lists the available ID types for the mapping
depending on the organism under investigation.
idTypes("hsa")
#> [1] "ACCNUM" "ALIAS" "ENSEMBL" "ENSEMBLPROT" "ENSEMBLTRANS"
#> [6] "ENTREZID" "ENZYME" "EVIDENCE" "EVIDENCEALL" "GENENAME"
#> [11] "GENETYPE" "GO" "GOALL" "IPI" "MAP"
#> [16] "OMIM" "ONTOLOGY" "ONTOLOGYALL" "PATH" "PFAM"
#> [21] "PMID" "PROSITE" "REFSEQ" "SYMBOL" "UCSCKG"
#> [26] "UNIPROT"ID mapping for the airway dataset (from ENSEMBL to ENTREZ gene ids) can
then be carried out using the function idMap.
head(rownames(airSE))
#> [1] "ENSG00000000003" "ENSG00000000419" "ENSG00000000457" "ENSG00000000460"
#> [5] "ENSG00000000971" "ENSG00000001036"
airSE <- idMap(airSE, org ="hsa", from = "ENSEMBL", to = "ENTREZID")
head(rownames(airSE))
#> [1] "7105" "8813" "57147" "55732" "3075" "2519"Now, we subject the ALL and the airway gene expression data to the enrichment analysis.
7 Enrichment analysis
In the following, we introduce how the EnrichmentBrowser package can be used to perform state-of-the-art enrichment analysis of gene sets. We consider the ALL and the airway gene expression data as processed in the previous sections. We are now interested in whether pre-defined sets of genes that are known to work together, e.g. as defined in the Gene Ontology (GO) or the KEGG pathway annotation, are coordinately differentially expressed.
7.1 Obtaining gene sets
The function getGenesets can be used to download gene sets from
databases such as GO and KEGG. We can use the function to download all
KEGG pathways for a chosen organism (here: Homo sapiens) as gene sets.
kegg.gs <- getGenesets(org = "hsa", db = "kegg")Analogously, the function getGenesets can be used to retrieve GO terms
of a selected ontology (here: biological process, BP) as defined in the
GO.db annotation package.
go.gs <- getGenesets(org = "hsa", db = "go", onto = "BP", mode = "GO.db")If provided a file, the function parses user-defined gene sets from GMT file format. Here, we use this functionality for reading a list of already downloaded KEGG gene sets for Homo sapiens containing NCBI Entrez Gene IDs.
gmt.file <- file.path(data.dir, "hsa_kegg_gs.gmt")
hsa.gs <- getGenesets(gmt.file)
length(hsa.gs)
#> [1] 39
hsa.gs[1:2]
#> $hsa05416_Viral_myocarditis
#> [1] "100509457" "101060835" "1525" "1604" "1605" "1756"
#> [7] "1981" "1982" "25" "2534" "27" "3105"
#> [13] "3106" "3107" "3108" "3109" "3111" "3112"
#> [19] "3113" "3115" "3117" "3118" "3119" "3122"
#> [25] "3123" "3125" "3126" "3127" "3133" "3134"
#> [31] "3135" "3383" "3683" "3689" "3908" "4624"
#> [37] "4625" "54205" "5551" "5879" "5880" "5881"
#> [43] "595" "60" "637" "6442" "6443" "6444"
#> [49] "6445" "71" "836" "841" "842" "857"
#> [55] "8672" "940" "941" "942" "958" "959"
#>
#> $`hsa04622_RIG-I-like_receptor_signaling_pathway`
#> [1] "10010" "1147" "1432" "1540" "1654" "23586" "26007" "29110"
#> [9] "338376" "340061" "3439" "3440" "3441" "3442" "3443" "3444"
#> [17] "3445" "3446" "3447" "3448" "3449" "3451" "3452" "3456"
#> [25] "3467" "3551" "3576" "3592" "3593" "3627" "3661" "3665"
#> [33] "4214" "4790" "4792" "4793" "5300" "54941" "55593" "5599"
#> [41] "5600" "5601" "5602" "5603" "56832" "57506" "5970" "6300"
#> [49] "64135" "64343" "6885" "7124" "7186" "7187" "7189" "7706"
#> [57] "79132" "79671" "80143" "841" "843" "8517" "8717" "8737"
#> [65] "8772" "9140" "9474" "9636" "9641" "9755"Note #1: Use getGenesets with db = "msigdb" to obtain gene set
collections for 11 different species from the Molecular Signatures
Database
(MSigDB).
Analogously, getGenesets with db = "enrichr" allows to obtain gene
set libraries from the comprehensive Enrichr
collection for 5 different
species.
Note #2: The idMap function can be used to map gene sets from NCBI
Entrez Gene IDs to other common gene ID types such as ENSEMBL gene IDs
or HGNC symbols as described in Section 6.
7.2 Set-based enrichment analysis
Currently, the following set-based enrichment methods are supported
sbeaMethods()
#> [1] "ora" "safe" "gsea" "gsa" "padog"
#> [6] "globaltest" "roast" "camera" "gsva" "samgs"
#> [11] "ebm" "mgsa"ORA: Overrepresentation Analysis (simple and frequently used test based on the hypergeometric distribution, see [4] for a critical review),
SAFE: Significance Analysis of Function and Expression (resampling version of ORA implements additional test statistics, e.g. Wilcoxon’s rank sum, and allows us to estimate the significance of gene sets by sample permutation; implemented in the safe package),
GSEA: Gene Set Enrichment Analysis (frequently used and widely accepted, uses a Kolmogorov–Smirnov statistic to test whether the ranks of the p-values of genes in a gene set resemble a uniform distribution [5]),
PADOG: Pathway Analysis with Down-weighting of Overlapping Genes (incorporates gene weights to favor genes appearing in few pathways versus genes that appear in many pathways; implemented in the PADOG package),
ROAST: ROtAtion gene Set Test (uses rotation instead of permutation for assessment of gene set significance; implemented in the limma and edgeR packages for microarray and RNA-seq data, respectively),
CAMERA: Correlation Adjusted MEan RAnk gene set test (accounts for inter-gene correlations as implemented in the limma and edgeR packages for microarray and RNA-seq data, respectively),
GSA: Gene Set Analysis (differs from GSEA by using the maxmean statistic, i.e. the mean of the positive or negative part of the gene scores in the gene set; implemented in the GSA package),
GSVA: Gene Set Variation Analysis (transforms the data from a gene by sample matrix to a gene set by sample matrix, thereby allowing the evaluation of gene set enrichment for each sample; implemented in the GSVA package)
GLOBALTEST: Global testing of groups of genes (general test of groups of genes for association with a response variable; implemented in the globaltest package),
SAMGS: Significance Analysis of Microarrays on Gene Sets (extending the SAM method for single genes to gene set analysis [6]),
EBM: Empirical Brown’s Method (combines p-values of genes in a gene set using Brown’s method to combine p-values from dependent tests; implemented in EmpiricalBrownsMethod),
MGSA: Model-based Gene Set Analysis (Bayesian modeling approach taking set overlap into account by working on all sets simultaneously, thereby reducing the number of redundant sets; implemented in mgsa).
See also Appendix 12 for a comprehensive introduction on underlying statistical concepts.
7.3 Guidelines for input and method selection
We recently performed a comprehensive assessment of the available set-based enrichment methods, and identified significant differences in runtime and applicability to RNA-seq data, the fraction of enriched gene sets depending on the null hypothesis tested, and the detection of relevant processes [7]. Based on these results, we make practical recommendations on how methods originally developed for microarray data can efficiently be applied to RNA-seq data, how to interpret results depending on the type of gene set test conducted and which methods are best suited to effectively prioritize gene sets with high phenotype relevance:
for the exploratory analysis of simple gene lists, we recommend ORA given its ease of applicability, fast runtime, and evident relevance of resulting gene set rankings provided that input gene list and reference gene list are chosen carefully and remembering ORA’s propensity for type I error rate inflation when genes tend to be co-expressed within sets.
for the analysis of pre-ranked gene lists accompanied by gene scores such as fold changes, alternatives to ORA such as pre-ranked GSEA or pre-ranked CAMERA exists.
for expression-based enrichment analysis on the full expression matrix, we recommend providing normalized log2 intensities for microarray data, and logTPMs (or logRPKMs / logFPKMs) for RNA-seq data. When given raw read counts, we recommend applying a variance-stabilizing transformation such as
voomto arrive at library-size normalized logCPMs.if the question of interest is to test for association of any gene in the set with the phenotype (self-contained null hypothesis), we recommend ROAST or GSVA that both test a directional hypothesis (genes in the set tend to be either predominantly up- or down-regulated). Both methods can be applied for simple or extended experimental designs, where ROAST is the more natural choice for the comparison of sample groups and also allows one to test a mixed hypothesis (genes in the set tend to be differentially expressed, regardless of the direction). The main strength of GSVA lies in its capabilities for analyzing single samples.
if the question of interest is to test for an excess differential expression in a gene set relative to genes outside the set (competitive null hypothesis), which we believe comes closest to the expectations and intuition of most end users when performing GSEA, we recommend PADOG, which is slower to run but resolves major shortcomings of ORA, and has desirable properties for the analyzed criteria and when compared to other competitive methods. However, PADOG is limited to testing a mixed hypothesis in a comparison of two sample groups, optionally including paired samples or sample batches. Therefore, we recommend the highly customizable SAFE for testing a directional hypothesis or in situations of more complex experimental designs such as comparisons between multiple groups, continuous phenotypes or the presence of covariates.
See also Reference [7] for the results of the benchmarking study and the GSEABenchmarkeR package for a general framework for reproducible benchmarking of gene set enrichment methods.
7.3.1 Microarray data
Given normalized log2 intensities for the ALL microarray dataset, a basic ORA can be carried out via
sbea.res <- sbea(method = "ora", se = allSE, gs = hsa.gs, perm = 0, alpha = 0.1)
gsRanking(sbea.res)
#> DataFrame with 4 rows and 4 columns
#> GENE.SET NR.GENES NR.SIG.GENES PVAL
#> <character> <numeric> <numeric> <numeric>
#> 1 hsa05130_Pathogenic_.. 44 5 0.0298
#> 2 hsa05206_MicroRNAs_i.. 133 10 0.0371
#> 3 hsa04622_RIG-I-like_.. 55 5 0.0681
#> 4 hsa04670_Leukocyte_t.. 94 7 0.0810Note that we set perm = 0 to invoke the classical hypergeometric test
without sample permutation, and that we chose a significance level
\(\alpha\) of 0.1 for demonstration purposes.
7.3.2 RNA-seq data
When analyzing RNA-seq datasets with expression values given as logTPMs
(or logRPKMs / logFPKMs), the available set-based enrichment methods can
be applied as for microarray data. However, when given raw read counts
as for the airway dataset, we recommend to first apply a
variance-stabilizing transformation such as voom to arrive at
library-size normalized logCPMs.
airSE <- normalize(airSE, norm.method = "vst")The mean-variance relationship of the transformed data is similar to what is observed for microarray data, simplifying the application of legacy enrichment methods such as GSEA and PADOG to RNA-seq data, and enable the use of fast and established methods.
air.res <- sbea(method = "gsea", se = airSE, gs = hsa.gs)
gsRanking(sbea.res)7.4 Result exploration
The result of every enrichment analysis is a ranking of gene sets by the
corresponding p-value. The gsRanking function displays by default
only those gene sets satisfying the chosen significance level \(\alpha\),
but we can use to obtain the full ranking.
gsRanking(sbea.res, signif.only = FALSE)
#> DataFrame with 39 rows and 4 columns
#> GENE.SET NR.GENES NR.SIG.GENES PVAL
#> <character> <numeric> <numeric> <numeric>
#> 1 hsa05130_Pathogenic_.. 44 5 0.0298
#> 2 hsa05206_MicroRNAs_i.. 133 10 0.0371
#> 3 hsa04622_RIG-I-like_.. 55 5 0.0681
#> 4 hsa04670_Leukocyte_t.. 94 7 0.0810
#> 5 hsa05100_Bacterial_i.. 64 5 0.1130
#> ... ... ... ... ...
#> 35 hsa05218_Melanoma 58 0 1
#> 36 hsa05150_Staphylococ.. 46 0 1
#> 37 hsa03420_Nucleotide_.. 41 0 1
#> 38 hsa03030_DNA_replica.. 33 0 1
#> 39 hsa03410_Base_excisi.. 27 0 1While such a ranked list is the standard output of existing enrichment
tools, the package EnrichmentBrowser provides
visualization and interactive exploration of resulting gene sets far
beyond that point. Using the eaBrowse function creates a HTML summary
from which each gene set can be inspected in detail (this builds on
functionality from the ReportingTools package).
The various options are described in Figure 1.
eaBrowse(sbea.res)
Figure 1: ORA result view For each significant gene set in the
ranking, the user can select to view (1) a gene report, that lists all
genes of a set along with fold change and DE p-value, (2) interactive
overview plots such as heatmap, p-value distribution, and volcano plot,
(3) the pathway in KEGG with differentially expressed genes highlighted
in red.
7.5 Network-based enrichment analysis
Having found sets of genes that are differentially regulated in the ALL data, we are now interested in whether these findings can be supported by known regulatory interactions.
For example, we want to know whether transcription factors and their target genes are expressed in accordance to the connecting regulations (activation/inhibition). Such information is usually given in a gene regulatory network derived from specific experiments or compiled from the literature ([8] for an example).
There are well-studied processes and organisms for which comprehensive
and well-annotated regulatory networks are available, e.g. the
RegulonDB for E. coli and Yeastract for S. cerevisiae. However,
there are also cases where such a network is missing. A basic workaround
is to compile a network from regulations in pathway databases such as
KEGG.
hsa.grn <- compileGRN(org="hsa", db="kegg")
head(hsa.grn)
#> FROM TO TYPE
#> [1,] "10000" "100132074" "-"
#> [2,] "10000" "1026" "+"
#> [3,] "10000" "1026" "-"
#> [4,] "10000" "1027" "-"
#> [5,] "10000" "10488" "+"
#> [6,] "10000" "107" "+"Now, we are able to perform enrichment analysis using the compiled network. Currently, the following network-based enrichment analysis methods are supported
nbeaMethods()
#> [1] "ggea" "spia" "pathnet" "degraph" "ganpa"
#> [6] "cepa" "topologygsa" "netgsa" "neat"GGEA: Gene Graph Enrichment Analysis (evaluates the consistency of known regulatory interactions with the observed expression data [9]),
SPIA: Signaling Pathway Impact Analysis (combines ORA with the the probability that expression changes are propagated across the pathway topology; implemented in the SPIA package),
PathNet: Pathway Analysis using Network Information (applies ORA on combined evidence for the observed signal for gene nodes and the signal implied by connected neighbors in the network; implemented in the PathNet package),
DEGraph: Differential expression testing for gene graphs (multivariate testing of differences in mean incorporating underlying graph structure; implemented in the DEGraph package),
TopologyGSA: Topology-based Gene Set Analysis (uses Gaussian graphical models to incorporate the dependence structure among genes as implied by pathway topology; implemented in the topologyGSA package),
GANPA: Gene Association Network-based Pathway Analysis (incorporates network-derived gene weights in the enrichment analysis; implemented in the GANPA package),
CePa: Centrality-based Pathway enrichment (incorporates network centralities as node weights mapped from differentially expressed genes in pathways; implemented in the CePa package),
NetGSA: Network-based Gene Set Analysis (incorporates external information about interactions among genes as well as novel interactions learned from data; implemented in the NetGSA package),
For demonstration, we perform GGEA using the compiled KEGG regulatory network.
nbea.res <- nbea(method="ggea", se=allSE, gs=hsa.gs, grn=hsa.grn)
gsRanking(nbea.res)
#> DataFrame with 9 rows and 5 columns
#> GENE.SET NR.RELS RAW.SCORE NORM.SCORE PVAL
#> <character> <numeric> <numeric> <numeric> <numeric>
#> 1 hsa04622_RIG-I-like_.. 36 13.70 0.382 0.000999
#> 2 hsa05416_Viral_myoca.. 7 3.29 0.470 0.002000
#> 3 hsa04520_Adherens_ju.. 12 4.86 0.405 0.005000
#> 4 hsa05217_Basal_cell_.. 17 6.60 0.388 0.008990
#> 5 hsa04390_Hippo_signa.. 66 23.20 0.351 0.013000
#> 6 hsa05134_Legionellosis 18 6.71 0.373 0.016000
#> 7 hsa05412_Arrhythmoge.. 5 2.14 0.429 0.025000
#> 8 hsa04621_NOD-like_re.. 40 13.90 0.349 0.025000
#> 9 hsa04210_Apoptosis 59 20.10 0.341 0.034000The resulting ranking lists, for each statistically significant gene
set, the number of relations of the network involving a member of the
gene set under study (NR.RELS), the sum of consistencies over the
relations of the set (RAW.SCORE), the score normalized by induced
network size (NORM.SCORE = RAW.SCORE / NR.RELS), and the
statistical significance of each gene set based on a permutation
approach.
A GGEA graph for a gene set depicts the consistency of each interaction in the set. Nodes (genes) are colored according to the expression (up-/down-regulated) and edges (interactions) are colored according to consistency, i.e. how well the interaction type (activation/inhibition) is reflected in the correlation of the observed expression of both interaction partners.
par(mfrow=c(1,2))
ggeaGraph( gs=hsa.gs[["hsa05217_Basal_cell_carcinoma"]], grn=hsa.grn, se=allSE)
ggeaGraphLegend()
7.6 User-defined enrichment methods
The goal of the EnrichmentBrowser package is to provide frequently used enrichment methods. However, it is also possible to exploit its visualization capabilities with user-defined set-based enrichment methods.
This requires implementing a function that takes the characteristic
arguments se (expression data) and gs (gene sets).
In addition, it must return a numeric vector ps storing the resulting
p-value for each gene set in gs. The p-value vector must also be
named accordingly (i.e.names(ps) == names(gs)).
Let us consider the following dummy enrichment method, which randomly renders five gene sets significant and the remaining insignificant.
dummySBEA <- function(se, gs) {
sig.ps <- sample(seq(0, 0.05, length = 1000), 5)
insig.ps <- sample(seq(0.1, 1, length = 1000), length(gs) - 5)
ps <- sample(c(sig.ps, insig.ps), length(gs))
names(ps) <- names(gs)
return(ps)
}We can plug this method into sbea as before.
sbea.res2 <- sbea(method = dummySBEA, se = allSE, gs = hsa.gs)
gsRanking(sbea.res2)
#> DataFrame with 5 rows and 2 columns
#> GENE.SET PVAL
#> <character> <numeric>
#> 1 hsa05410_Hypertrophi.. 0.0179
#> 2 hsa00790_Folate_bios.. 0.0229
#> 3 hsa04514_Cell_adhesi.. 0.0396
#> 4 hsa04350_TGF-beta_si.. 0.0428
#> 5 hsa03420_Nucleotide_.. 0.0434As described in the previous section, it is also possible to analogously
plugin user-defined network-based enrichment methods into nbea.
8 Combining results
Different enrichment analysis methods usually result in different gene
set rankings for the same dataset. To compare results and detect gene
sets that are supported by different methods, the
EnrichmentBrowser package allows combining results from
the different set-based and network-based enrichment analysis methods.
The combination of results yields a new ranking of the gene sets under
investigation by specified ranking criteria, e.g. the average rank
across methods. We consider the ORA result and the GGEA result from the
previous sections and use the function combResults.
res.list <- list(sbea.res, nbea.res)
comb.res <- combResults(res.list)The combined result can be detailedly inspected as before and interactively ranked as depicted in Figure 2.
eaBrowse(comb.res, graph.view=hsa.grn, nr.show=5)
Figure 2: Combined result view By clicking on one of the
columns (ORA.RANK, …, GGEA.PVAL) the result can be interactively
ranked according to the selected
criterion.
9 Putting it all together
There are cases where it is necessary to perform certain steps of the
demonstrated enrichment analysis pipeline individually. However, it is
often more convenient to run the complete standardized pipeline. This
can be done using the all-in-one wrapper function ebrowser. For
example, the result page displayed in Figure 2 can
also be produced from scratch via
ebrowser( meth=c("ora", "ggea"), exprs=exprs.file, cdat=cdat.file, rdat=rdat.file, org="hsa", gs=hsa.gs, grn=hsa.grn, comb=TRUE, nr.show=5)10 Advanced: configuration parameters
Similar to R’s options settings, the EnrichmentBrowser
uses certain package-wide configuration parameters, which affect the way
in which analysis is carried out and how results are displayed. The
settings of these parameters can be examined and, to some extent, also
changed using the function configEBrowser. For instance, the default
directory where the EnrichmentBrowser writes results to
can be updated via
configEBrowser(key="OUTDIR.DEFAULT", value="/my/out/dir")and examined via
configEBrowser("OUTDIR.DEFAULT")
#> [1] "/my/out/dir"Note that changing these defaults should be done with care, as inappropriate settings might impair the package’s functionality. The complete list of incorporated configuration parameters along with their default settings can be inspected via
?configEBrowser11 For non-R users: command line invocation
The package source contains two scripts in inst/scripts to invoke the
EnrichmentBrowser from the command line using Rscript.
The de_rseq.R script is a lightweight wrapper script to carry out
differential expression analysis of RNA-seq data either based on
limma (using the voom-transformation),
edgeR, or DESeq2.
The eBrowserCMD.R implements the full functionality and allows to
carry out the various enrichment methods and to produce HTML reports for
interactive exploration of results.
The inst/scripts folder also contains a README file that
comprehensively documents the usage of both scripts.
12 A primer on terminology and statistical theory
12.1 Where does it all come from?
Test whether known biological functions or processes are over-represented (= enriched) in an experimentally-derived gene list, e.g. a list of differentially expressed (DE) genes. See [4] for a critical review.
Example: Transcriptomic study, in which 12,671 genes have been tested for differential expression between two sample conditions and 529 genes were found DE.
Among the DE genes, 28 are annotated to a specific functional gene set, which contains in total 170 genes. This setup corresponds to a \(2\times2\) contingency table,
deTable <- matrix(c(28, 142, 501, 12000),
nrow = 2,
dimnames = list(c("DE", "Not.DE"), c("In.gene.set", "Not.in.gene.set")))
deTable
#> In.gene.set Not.in.gene.set
#> DE 28 501
#> Not.DE 142 12000where the overlap of 28 genes can be assessed based on the hypergeometric distribution. This corresponds to a one-sided version of Fisher’s exact test, yielding here a significant enrichment.
fisher.test(deTable, alternative = "greater")
#>
#> Fisher's Exact Test for Count Data
#>
#> data: deTable
#> p-value = 4.088e-10
#> alternative hypothesis: true odds ratio is greater than 1
#> 95 percent confidence interval:
#> 3.226736 Inf
#> sample estimates:
#> odds ratio
#> 4.721744This basic principle is at the foundation of major public and commercial enrichment tools such as DAVID and Pathway Studio.
Although gene set enrichment methods have been primarily developed and applied on transcriptomic data, they have recently been modified, extended and applied also in other fields of genomic and biomedical research. This includes novel approaches for functional enrichment analysis of proteomic and metabolomic data as well as genomic regions and disease phenotypes [10–13].
12.2 Gene sets, pathways & regulatory networks
Gene sets are simple lists of usually functionally related genes without further specification of relationships between genes.
Pathways can be interpreted as specific gene sets, typically representing a group of genes that work together in a biological process. Pathways are commonly divided in metabolic and signaling pathways. Metabolic pathways such as glycolysis represent biochemical substrate conversions by specific enzymes. Signaling pathways such as the MAPK signaling pathway describe signal transduction cascades from receptor proteins to transcription factors, resulting in activation or inhibition of specific target genes.
Gene regulatory networks describe the interplay and effects of regulatory factors (such as transcription factors and microRNAs) on the expression of their target genes.
12.3 Resources
GO and KEGG annotations are most frequently used for the enrichment analysis of functional gene sets. Despite an increasing number of gene set and pathway databases, they are typically the first choice due to their long-standing curation and availability for a wide range of species.
Gene Ontology (GO) consists of three major sub-ontologies that classify gene products according to molecular function (MF), biological process (BP), and cellular component (CC). Each ontology consists of GO terms that define MFs, BPs, or CCs to which specific genes are annotated. The terms are organized in a directed acyclic graph, where the edges between the terms represent relationships of different types. They relate the terms according to a parent-child scheme, i.e. parent terms denote more general entities, whereas child terms represent more specific entities.
The Kyoto Encyclopedia of Genes and Genomes (KEGG) is a collection of manually drawn pathway maps representing molecular interaction and reaction networks. These pathways cover a wide range of biochemical processes that can be divided into 7 broad categories: metabolism, genetic and environmental information processing, cellular processes, organismal systems, human diseases, and drug development. Metabolism and drug development pathways differ from pathways of the other 5 categories by illustrating reactions between chemical compounds. Pathways of the other 5 categories illustrate molecular interactions between genes and gene products.
12.4 Gene set analysis vs. gene set enrichment analysis
The two predominantly used enrichment methods are:
Overrepresentation analysis (ORA), testing whether a gene set contains disproportional many genes of significant expression change, based on the procedure outlined in section 12.1,
Gene set enrichment analysis (GSEA), testing whether genes of a gene set accumulate at the top or bottom of the full gene vector ordered by direction and magnitude of expression change [5].
However, the term gene set enrichment analysis nowadays subsumes a general strategy implemented by a wide range of methods [14]. Those methods have in common the same goal, although the approach and statistical model can vary substantially [4, 15].
To better distinguish from the specific method, some authors use the term gene set analysis to denote the general strategy. However, there is also a specific method of this name [16].
12.5 Underlying null: competitive vs. self-contained
Goeman and Buehlmann, 2007, classified existing enrichment methods into competitive and self-contained based on the underlying null hypothesis [4].
Competitive null hypothesis: the genes in the set of interest are at most as often DE as the genes not in the set,
Self-contained null hypothesis: no genes in the set of interest are DE.
Although the authors argue that a self-contained null is closer to the the actual question of interest, the vast majority of enrichment methods is competitive.
Goeman and Buehlmann further raise several critical issues concerning the \(2\times2\) ORA:
a rather arbitrary classification of genes in DE / not DE,
based on gene sampling, although sampling of subjects is appropriate,
unrealistic independence assumption between genes, resulting in highly anti-conservative p-values.
With regard to these statistical concerns, GSEA is considered superior:
takes all measured genes into account,
subject sampling via permutation of class labels,
the incorporated permutation procedure implicitly accounts for correlations between genes.
However, the simplicity and general applicability of ORA is unmet by subsequent methods improving on these issues. For instance, GSEA requires the expression data as input, which is not available for gene lists derived from other experiment types. On the other hand, the involved sample permutation procedure has been proven inaccurate and time-consuming [16–18].
12.6 Generations: ora, fcs & topology-based
Khatri et al., 2012, have taken a slightly different approach by classifying methods along the timeline of development into three generations [15]:
Generation: ORA methods based on the \(2\times2\) contingency table test,
Generation: functional class scoring (FCS) methods such as GSEA, which compute gene set (= functional class) scores by summarizing per-gene DE statistics,
Generation: topology-based methods, explicitly taking into account interactions between genes as defined in signaling pathways and gene regulatory networks ([9] for an example).
Although topology-based (also: network-based) methods appear to be the most realistic, their straightforward application can be impaired by features that are not detectable on the transcriptional level (such as protein-protein interactions) and insufficient network knowledge [8, 19].
Given the individual benefits and limitations of existing methods, cautious interpretation of results is required to derive valid conclusions. Whereas no single method is best suited for all application scenarios, applying multiple methods can be beneficial. This has been shown to filter out spurious hits of individual methods, thereby reducing the outcome to gene sets accumulating evidence from different methods [20, 21].
13 Frequently asked questions
How to cite the EnrichmentBrowser ? Geistlinger L, Csaba G and Zimmer R. Bioconductor’s EnrichmentBrowser: seamless navigation through combined results of set- & network-based enrichment analysis. BMC Bioinformatics, 17:45, 2016.
Is it possible to apply the EnrichmentBrowser to simple gene lists? Enrichment methods implemented in the EnrichmentBrowser are, except for ORA, expression-based (and also draw their strength from that). The set-based methods GSEA, SAFE, and SAMGS use sample permutation, involving recomputation of differential expression, for gene set significance estimation, i.e. they require the complete expression matrix. The network-based methods require measures of differential expression such as fold change and p-value to score interactions of the network. In addition, visualization of enriched gene sets is explicitly designed for expression data. Thus, for simple gene list enrichment, tools like DAVID and GeneAnalytics are more suitable, and it is recommended to use them for this purpose.
References
1. Chiaretti S, Li X, Gentleman R, Vitale A, Vignetti M, Mandelli F, et al. Gene expression profile of adult t-cell acute lymphocytic leukemia identifies distinct subsets of patients with different response to therapy and survival. Blood. 2004;103:2771–8.
2. Gentleman R, Carey VJ, Huber W, Irizarry RA, Dudoit S, editors. Bioinformatics and computational biology solutions using r and bioconductor. Springer New York; 2005.
3. Himes BE, Jiang X, Wagner P, Hu R, Wang Q, Klanderman B, et al. RNA-seq transcriptome profiling identifies crispld2 as a glucocorticoid responsive gene that modulates cytokine function in airway smooth muscle cells. PLoS ONE. 2014;9:e99625.
4. Goeman JJ, Bühlmann P. Analyzing gene expression data in terms of gene sets: Methodological issues. Bioinformatics. 2007;23:980–7.
5. Subramanian A, Tamayo P, Mootha VK, Mukherjee S, Ebert BL, Gillette MA, et al. Gene set enrichment analysis: A knowledge-based approach for interpreting genome-wide expression profiles. Proceedings of the National Academy of Sciences. 2005;102:15545–50.
6. Dinu I, Potter JD, Mueller T, Liu Q, Adewale AJ, Jhangri GS, et al. Improving gene set analysis of microarray data by sam-gs. BMC Bioinformatics. 2007;8.
7. Geistlinger L, Csaba G, Santarelli M, Ramos M, Schiffer L, Turaga N, et al. Toward a gold standard for benchmarking gene set enrichment analysis. Briefings in Bioinformatics. 2020;22:545–56.
8. Geistlinger L, Csaba G, Dirmeier S, Kuffner R, Zimmer R. A comprehensive gene regulatory network for the diauxic shift in saccharomyces cerevisiae. Nucleic Acids Research. 2013;41:8452–63.
9. Geistlinger L, Csaba G, Küffner R, Mulder N, Zimmer R. From sets to graphs: Towards a realistic enrichment analysis of transcriptomic systems. Bioinformatics. 2011;27:i366–73.
10. Lavallée-Adam M, Yates JR. Using psea-Quant for protein set enrichment analysis of quantitative mass spectrometry-Based proteomics. Current Protocols in Bioinformatics. 2016;53.
11. Chagoyen M, López-Ibáñez J, Pazos F. Functional analysis of metabolomics data. In. Springer New York; 2016. pp. 399–406.
12. McLean CY, Bristor D, Hiller M, Clarke SL, Schaar BT, Lowe CB, et al. GREAT improves functional interpretation of cis-regulatory regions. Nature Biotechnology. 2010;28:495–501.
13. Ried JS, Döring A, Oexle K, Meisinger C, Winkelmann J, Klopp N, et al. PSEA: Phenotype set enrichment analysis-a new method for analysis of multiple phenotypes. Genetic Epidemiology. 2012;36:244–52.
14. Huang DW, Sherman BT, Lempicki RA. Bioinformatics enrichment tools: Paths toward the comprehensive functional analysis of large gene lists. Nucleic Acids Research. 2008;37:1–13.
15. Khatri P, Sirota M, Butte AJ. Ten years of pathway analysis: Current approaches and outstanding challenges. PLoS Computational Biology. 2012;8:e1002375.
16. Efron B, Tibshirani R. On testing the significance of sets of genes. The Annals of Applied Statistics. 2007;1.
17. Phipson B, Smyth GK. Permutation p-values should never be zero: Calculating exact p-values when permutations are randomly drawn. Statistical Applications in Genetics and Molecular Biology. 2010;9.
18. Larson JL, Owen AB. Moment based gene set tests. BMC Bioinformatics. 2015;16.
19. Bayerlová M, Jung K, Kramer F, Klemm F, Bleckmann A, Beißbarth T. Comparative study on gene set and pathway topology-based enrichment methods. BMC Bioinformatics. 2015;16.
20. Geistlinger L, Csaba G, Zimmer R. Bioconductor’s enrichmentbrowser: Seamless navigation through combined results of set- & network-based enrichment analysis. BMC Bioinformatics. 2016;17.
21. Alhamdoosh M, Ng M, Wilson NJ, Sheridan JM, Huynh H, Wilson MJ, et al. Combining multiple tools outperforms individual methods in gene set enrichment analyses. Bioinformatics. 2016;33:414–24.