Enrichment Analysis and Visualization
8 July 2019
Contents
1 Introduction
InterMine constitutes a powerful open source data warehouse system which integrates diverse biological data sets and provides a growing number of analysis tools, including enrichment analysis widgets (Smith et al. 2012; Kalderimis et al. 2014).
Specifically, the gene set enrichment tool looks for annotations to a set of genes that occur more than would be expected by chance, given a background population of annotations. The hypergeometric distribution is used to calculate a P-value for each annotation and a choice of correction methods for multiple testing (Bonferonni, Holm-Bonferonni and Benjamini Hochberg) are available (Smith et al. 2012; Kalderimis et al. 2014).
InterMine provides Gene Ontology enrichment statistics as well as enrichment statistics for other annotation types including protein domains, pathways, human diseases, mammalian phenotypes and publications. The default background population is all genes in the genome with the specified annotation type. However, the background population can be changed by specifying another list. More information can be found in the online documentation.
2 Perform enrichment analysis with InterMineR
Performing enrichment analysis with InterMineR is preceded by two steps:
Retrieve the list of bioentities of interest (Genes, Proteins, SNPs, etc.) for which the analysis will be performed.
Get the enrichment widget name which indicates the annotation types that you want to investigate for enrichment (e.g. Gene Ontology Terms, protein domains, KEGG and Reactome Pathways, human diseases, mammalian phenotypes and publications).
2.1 Retrieve list containing the genes of interest from InterMineR
In this example, enrichment analysis is performed using a list of genes which are associated with all forms of Diabetes according to OMIM. All query results and calculations are correct according to HumanMine release 5.1 (November 2018). Genomic coordinates and p-values are likely to change between database releases but the methods will remain the same.
The PL_DiabetesGenes dataset is included in the InterMineR package and can also be found online in HumanMine.
library(InterMineR)Retrieve PL_DiabetesGenes data and get gene identifiers as HumanrMine Gene.symbol and Gene.primaryIdentifier values (ENTREZ IDs).
# get HumanMine instance
im.human = initInterMine(listMines()["HumanMine"])
# retrieve data model for HumanMine
# hsa_model = getModel(im.human) # temporarily removed
# all targets of HumanMine enrichment widgets possess InterMine ids
# subset(hsa_model, type %in% c("Gene", "Protein", "SNP") & child_name == "id") # temporarily removed
# load human genes which are associated with Diabetes (retrieved from HumanMine)
data("PL_DiabetesGenes")
head(PL_DiabetesGenes, 3)## Gene.symbol Gene.name
## 1 ABCC8 ATP binding cassette subfamily C member 8
## 2 ACE angiotensin I converting enzyme
## 3 AKT2 AKT serine/threonine kinase 2
## Gene.primaryIdentifier Gene.secondaryIdentifier Gene.length
## 1 6833 ENSG00000006071 83961
## 2 1636 ENSG00000159640 21320
## 3 208 ENSG00000105221 55215
## Gene.organism.name
## 1 Homo sapiens
## 2 Homo sapiens
## 3 Homo sapiens# get Gene.symbol
hsa_gene_names = as.character(PL_DiabetesGenes$Gene.symbol)
head(hsa_gene_names, 3)## [1] "ABCC8" "ACE" "AKT2"# get Gene.primaryIdentifiers (ENTREZ Gene identifiers)
hsa_gene_entrez = as.character(PL_DiabetesGenes$Gene.primaryIdentifier)
head(hsa_gene_entrez, 3)## [1] "6833" "1636" "208"2.2 Get enrichment widgets
After obtaining the Gene.id values of interest, we must determine the type of annotation for the enrichment analysis.
To retrieve all available widgets for a Mine we can use the getWidgets function.
# get all HumanMine widgets
human.widgets = as.data.frame(getWidgets(im.human))
human.widgets## startClassDisplay
## 1 primaryIdentifier
## 2 <NA>
## 3 primaryIdentifier
## 4 primaryIdentifier
## 5 <NA>
## 6 primaryIdentifier
## 7 primaryIdentifier
## 8 primaryIdentifier
## 9 primaryIdentifier
## 10 <NA>
## enrichIdentifier
## 1 GWASResults.study.publication.pubMedId
## 2 <NA>
## 3 goAnnotation.ontologyTerm.parents.identifier
## 4 publications.pubMedId
## 5 <NA>
## 6 pathways.identifier
## 7 GWASResults.study.publication.pubMedId
## 8 proteins.proteinDomainRegions.proteinDomain.primaryIdentifier
## 9 proteinDomainRegions.proteinDomain.primaryIdentifier
## 10 <NA>
## name
## 1 snp_gwas_study_enrichment
## 2 chromosome_distribution_for_snp
## 3 go_enrichment_for_gene
## 4 publication_enrichment
## 5 chromosome_distribution_for_gene
## 6 pathway_enrichment
## 7 snp_publication_enrichment
## 8 prot_dom_enrichment_for_gene
## 9 prot_dom_enrichment_for_protein
## 10 interactions
## description
## 1 GWAS studies enriched for SNPs in this list.
## 2 Actual: number of items in this list found on each chromosome. Expected: given the total number of items on the chromosome and the number of items in this list, the number of items expected to be found on each chromosome.
## 3 GO terms enriched for items in this list.
## 4 Publications enriched for genes in this list.
## 5 Actual: number of items in this list found on each chromosome. Expected: given the total number of items on the chromosome and the number of items in this list, the number of items expected to be found on each chromosome.
## 6 Pathways enriched for genes in this list - data from KEGG and Reactome
## 7 Publications enriched for SNPs in this list.
## 8 Protein Domains enriched for items in this list.
## 9 Protein Domains enriched for items in this list.
## 10 Genes (from the list or not) that interact with genes in this list. Counts may include the same interaction more than once if observed in multiple experiments.
## enrich startClass
## 1 GWASResults.study.name SNP
## 2 <NA> SNP
## 3 goAnnotation.ontologyTerm.parents.name Gene
## 4 publications.title Gene
## 5 <NA> Gene
## 6 pathways.name Gene
## 7 GWASResults.study.publication.title SNP
## 8 proteins.proteinDomainRegions.proteinDomain.name Gene
## 9 proteinDomainRegions.proteinDomain.name Protein
## 10 <NA> <NA>
## title targets widgetType chartType
## 1 GWAS study Enrichment for SNPs SNP enrichment <NA>
## 2 Chromosome Distribution SNP chart ColumnChart
## 3 Gene Ontology Enrichment Gene enrichment <NA>
## 4 Publication Enrichment Gene enrichment <NA>
## 5 Chromosome Distribution Gene chart ColumnChart
## 6 Pathway Enrichment Gene enrichment <NA>
## 7 Publication Enrichment for SNPs SNP enrichment <NA>
## 8 Protein Domain Enrichment Gene enrichment <NA>
## 9 Protein Domain Enrichment Protein enrichment <NA>
## 10 Interactions Gene table <NA>
## filters
## 1 <NA>
## 2 organism.name=[list]
## 3 biological_process,cellular_component,molecular_function
## 4 <NA>
## 5 organism.name=[list]
## 6 All,KEGG pathways data set,Reactome data set
## 7 <NA>
## 8 <NA>
## 9 <NA>
## 10 <NA>
## labels
## 1 <NA>
## 2 Chromosome & Count
## 3 <NA>
## 4 <NA>
## 5 Chromosome & Count
## 6 <NA>
## 7 <NA>
## 8 <NA>
## 9 <NA>
## 10 <NA>HumanMine provides enrichment for genes, proteins and SNPs, but here we are interested only in the gene-related, enrichment widgets.
# get enrichment, gene-related widgets for HumanMine
subset(human.widgets, widgetType == 'enrichment' & targets == "Gene")## startClassDisplay
## 3 primaryIdentifier
## 4 primaryIdentifier
## 6 primaryIdentifier
## 8 primaryIdentifier
## enrichIdentifier
## 3 goAnnotation.ontologyTerm.parents.identifier
## 4 publications.pubMedId
## 6 pathways.identifier
## 8 proteins.proteinDomainRegions.proteinDomain.primaryIdentifier
## name
## 3 go_enrichment_for_gene
## 4 publication_enrichment
## 6 pathway_enrichment
## 8 prot_dom_enrichment_for_gene
## description
## 3 GO terms enriched for items in this list.
## 4 Publications enriched for genes in this list.
## 6 Pathways enriched for genes in this list - data from KEGG and Reactome
## 8 Protein Domains enriched for items in this list.
## enrich startClass
## 3 goAnnotation.ontologyTerm.parents.name Gene
## 4 publications.title Gene
## 6 pathways.name Gene
## 8 proteins.proteinDomainRegions.proteinDomain.name Gene
## title targets widgetType chartType
## 3 Gene Ontology Enrichment Gene enrichment <NA>
## 4 Publication Enrichment Gene enrichment <NA>
## 6 Pathway Enrichment Gene enrichment <NA>
## 8 Protein Domain Enrichment Gene enrichment <NA>
## filters labels
## 3 biological_process,cellular_component,molecular_function <NA>
## 4 <NA> <NA>
## 6 All,KEGG pathways data set,Reactome data set <NA>
## 8 <NA> <NA>The column names provides the character strings that can be passed as arguments to the doEnrichment function and thus define the type of enrichment analysis.
2.3 Perform enrichment analysis with InterMineR
We will perform Gene Ontology Enrichment analysis for the genes in our list using the Gene.id values and the ‘go_enrichment_for_gene’ widget.
# Assign directly the doEnrichment.string from getGeneIds function to the ids argument of doEnrichment to perform the analysis
# Perform enrichment analysis
GO_enrichResult = doEnrichment(
im = im.human,
ids = hsa_gene_entrez,
widget = "go_enrichment_for_gene"
# organism = "Homo sapiens" # optional if results from more than one organism are retrieved
)As mentioned above ‘PL_DiabetesGenes’ constitutes a genelist which already exists in HumanMine. Therefore, the enrichment analysis could also be performed by passing the name of the list to the genelist argument, with the same results:
# Perform enrichment analysis with genelist name instead of ids
GO_enrichResult = doEnrichment(
im = im.human,
genelist = "PL_DiabetesGenes",
# ids = hsa_gene_entrez,
widget = "go_enrichment_for_gene"
)doEnrichment function returns a list which contains:
- a data.frame with the results of the enrichment analysis which was performed in the defined InterMine platform. p-values are given after applying the correction algorithm. Count and populationAnnotationCount columns contain the number of genes that belong to each GO term in the list and in the background population respectively.
# data.frame containing the results of the enrichment analysis
head(GO_enrichResult$data)## identifier description pValue count
## 1 GO:0030072 peptide hormone secretion 4.336648956166819E-19 21
## 2 GO:0046879 hormone secretion 7.6663597110215735E-19 22
## 3 GO:0009914 hormone transport 9.540238337303802E-19 22
## 4 GO:0030073 insulin secretion 1.1561118844984925E-16 18
## 5 GO:0033500 carbohydrate homeostasis 2.8899958514686346E-16 17
## 6 GO:0042593 glucose homeostasis 3.1522069422433323E-16 17
## populationAnnotationCount
## 1 251
## 2 313
## 3 322
## 4 207
## 5 182
## 6 181dim(GO_enrichResult$data)## [1] 636 5- the size of the reference background population (populationCount)
GO_enrichResult$populationCount## [1] 16597- the number of input features that were not included in the enrichment analysis (notAnalyzed)
GO_enrichResult$notAnalysed## [1] 6- the name and url of the Mine (im)
GO_enrichResult$im## $mine
## HumanMine
## "http://www.humanmine.org/humanmine"
##
## $token
## [1] ""- the rest of the parameters used to perform the analysis (parameters)
GO_enrichResult$parameters## ids
## "1018504,1069788,1088766,1074405,1167250,1169335,1094079,1035007,1150284,1046584,1074688,1085117,1076201,1251964,1248044,1254789,1265202,1270014,1120348,1043788,1229453,1146398,1204931,1227904,1208398,1128029,1113856,1233700,1229612,1276310,1109526,1001868,1179209,1174712,1186790,1126059,1042350,1214384,1270234,1095332,1070530,1153906,1247509,1072698,1161160,1014965,1172352,1223354,1082710,1285639,1111701,1072131,1283829,1007739,1257384,1195770,1252086,1180051,1198453,1061004,1211404,1086036,1229758,1196726,1115541,1260234,1053657,1059762"
## widget
## "go_enrichment_for_gene"
## maxp
## "0.05"
## correction
## "Benjamini Hochberg"2.4 Apply filters on enrichment analysis
Some of the enrichment widgets contain filters which, when applied, limit the annotation types of the enrichment analysis.
In our example, if we want to retrieve only the enriched GO terms in our list of genes that are related to molecular function, we will use the appropriate filter:
# get available filter values for Gene Ontology Enrichment widget
as.character(subset(human.widgets, name == "go_enrichment_for_gene")$filters)## [1] "biological_process,cellular_component,molecular_function"# Perform enrichment analysis for GO molecular function terms
GO_MF_enrichResult = doEnrichment(
im = im.human,
ids = hsa_gene_entrez,
widget = "go_enrichment_for_gene",
filter = "molecular_function")
head(GO_MF_enrichResult$data)## identifier description
## 1 GO:0005102 signaling receptor binding
## 2 GO:0005158 insulin receptor binding
## 3 GO:0005159 insulin-like growth factor receptor binding
## 4 GO:0140297 DNA-binding transcription factor binding
## 5 GO:0001067 regulatory region nucleic acid binding
## 6 GO:0044212 transcription regulatory region DNA binding
## pValue count populationAnnotationCount
## 1 0.0037534905026989592 19 1545
## 2 0.004425288799249525 4 21
## 3 0.023859181561416545 3 16
## 4 0.023910794938533403 8 336
## 5 0.025815502329076468 13 911
## 6 0.033633006777797264 13 909dim(GO_MF_enrichResult$data)## [1] 6 52.5 Apply multiple test correction
To reduce the probability of false positive errors, three different multiple correction algorithms can be applied to the results of the enrichment analysis:
- Benjamini Hochberg (default)
- Holm-Bonferroni
- Bonferroni
The application of one of these algorithms changes the p-values and determines the number of the results which will be returned from the analysis:
# Perform enrichment analysis for Protein Domains in list of genes
PD_FDR_enrichResult = doEnrichment(
im = im.human,
ids = hsa_gene_entrez,
widget = "prot_dom_enrichment_for_gene",
correction = "Benjamini Hochberg"
)
head(PD_FDR_enrichResult$data)## NULL# Perform enrichment analysis for Protein Domains in list of genes
# but without a correction algoritm this time
PD_None_enrichResult = doEnrichment(
im = im.human,
ids = hsa_gene_entrez,
widget = "prot_dom_enrichment_for_gene",
correction = "None"
)
head(PD_None_enrichResult$data)## identifier description
## 1 IPR006897 Hepatocyte nuclear factor 1, beta isoform, C-terminal
## 2 IPR023219 Hepatocyte nuclear factor 1, N-terminal domain superfamily
## 3 IPR039066 Hepatocyte nuclear factor 1
## 4 IPR006899 Hepatocyte nuclear factor 1, N-terminal
## 5 IPR039011 Insulin receptor substrate
## 6 IPR006020 PTB/PI domain
## pValue count populationAnnotationCount
## 1 1.0948063262382918E-5 2 2
## 2 1.0948063262382918E-5 2 2
## 3 1.0948063262382918E-5 2 2
## 4 3.277347110668014E-5 2 3
## 5 3.277347110668014E-5 2 3
## 6 3.198964753864943E-4 3 403 Visualization of InterMineR Gene Ontology Enrichment analysis results
3.1 Convert InterMineR enrichment analysis results to a GeneAnswer-class object
In order to visualize the InterMineR enrichment analysis results, we will use the function convertToGeneAnswers. This function was created to facilitate the visualization of doEnrichment function results by converting them into a GeneAnswers-class object.
This way we can utilize the functions of the package GeneAnswers to visualize the results of the enrichment analysis and the relations between the features (e.g. genes) and/or the annoatation categories (e.g. GO terms) (Feng et al. 2010, 2012; Huang et al. 2014).
# load GeneAnswers package
library(GeneAnswers)
# convert InterMineR Gene Ontology Enrichment analysis results to GeneAnswers object
geneanswer_object = convertToGeneAnswers(
# assign with doEnrichment result:
enrichmentResult = GO_enrichResult,
# assign with list of genes:
geneInput = data.frame(GeneID = as.character(hsa_gene_entrez),
stringsAsFactors = FALSE),
# assign with the type of gene identifier
# in our example we use Gene.primaryIdentifier values (ENTREZ IDs):
geneInputType = "Gene.primaryIdentifier",
# assign with Bioconductor annotation package:
annLib = 'org.Hs.eg.db',
# assign with annotation category type
# in our example we use Gene Ontology (GO) terms:
categoryType = "GO"
#optional define manually if 'enrichIdentifier' is missing from getWidgets:
#enrichCategoryChildName = "Gene.goAnnotation.ontologyTerm.parents.identifier"
)
class(geneanswer_object)## [1] "GeneAnswers"
## attr(,"package")
## [1] "GeneAnswers"summary(geneanswer_object)## This GeneAnswers instance was build from GO based on hyperG test.
## Statistical information of 636 categories with p value less than 0.05 are reported. Other categories are considered as nonsignificant.
## There are 636 categories related to the given 68 genes
##
## Summary of GeneAnswers instance information:
##
## Slot: geneInput
## GeneID
## 1 6833
## 2 1636
## 3 208
## 4 26060
## 5 359
## 6 551
## ......
##
## Slot: testType
## [1] "hyperG"
##
## Slot: pvalueT
## [1] 0.05
##
## Slot: genesInCategory
## $`GO:0030072`
## [1] "6833" "640" "11132" "2645" "3077" "6927" "6928" "3172"
## [9] "3557" "3569" "3630" "3667" "8660" "3710" "3767" "4544"
## [17] "4760" "3651" "6514" "169026" "6934"
##
## $`GO:0046879`
## [1] "6833" "640" "11132" "2645" "3077" "6927" "6928" "3172"
## [9] "3557" "3569" "3630" "3667" "8660" "3710" "3767" "4544"
## [17] "4760" "3651" "56729" "6514" "169026" "6934"
##
## $`GO:0009914`
## [1] "6833" "640" "11132" "2645" "3077" "6927" "6928" "3172"
## [9] "3557" "3569" "3630" "3667" "8660" "3710" "3767" "4544"
## [17] "4760" "3651" "56729" "6514" "169026" "6934"
##
## $`GO:0030073`
## [1] "6833" "640" "11132" "2645" "6927" "6928" "3172" "3557"
## [9] "3667" "8660" "3710" "3767" "4544" "4760" "3651" "6514"
## [17] "169026" "6934"
##
## $`GO:0033500`
## [1] "6833" "2642" "2645" "6927" "3172" "3630" "3643" "3667"
## [9] "8660" "3767" "4544" "4760" "4938" "3651" "5468" "169026"
## [17] "6934" "7466"
##
## $`GO:0042593`
## [1] "6833" "2642" "2645" "6927" "3172" "3630" "3643" "3667"
## [9] "8660" "3767" "4544" "4760" "4938" "3651" "5468" "169026"
## [17] "6934" "7466"
##
## ......
##
## Slot: geneExprProfile
## NULL
##
## Slot: annLib
## [1] "org.Hs.eg.db"
##
## Slot: categoryType
## [1] "GO"
##
## Slot: enrichmentInfo
## genes in Category percent in the observed List
## GO:0030072 21 0.3088235
## GO:0046879 22 0.3235294
## GO:0009914 22 0.3235294
## GO:0030073 18 0.2647059
## GO:0033500 17 0.2500000
## GO:0042593 17 0.2500000
## percent in the genome fold of overrepresents odds ratio
## GO:0030072 0.01512322 20.42049 29.09774
## GO:0046879 0.01885883 17.15533 24.88179
## GO:0009914 0.01940110 16.67583 24.17297
## GO:0030073 0.01247213 21.22379 28.50435
## GO:0033500 0.01096584 22.79808 30.06410
## GO:0042593 0.01090559 22.92403 30.23204
## p value fdr p value
## GO:0030072 4.336649e-19 4.336649e-19
## GO:0046879 7.666360e-19 7.666360e-19
## GO:0009914 9.540238e-19 9.540238e-19
## GO:0030073 1.156112e-16 1.156112e-16
## GO:0033500 2.889996e-16 2.889996e-16
## GO:0042593 3.152207e-16 3.152207e-16
## ......#slotNames(geneanswer_object)GeneAnswers is package designed for the enrichment analysis and visualization of gene lists. It is worth mentioning that the InterMineR filters for the widgets of Gene Ontology and Pathway Enrichment:
as.character(
subset(human.widgets,
title %in% c("Gene Ontology Enrichment",
"Pathway Enrichment"))$filters
)## [1] "biological_process,cellular_component,molecular_function"
## [2] "All,KEGG pathways data set,Reactome data set"match the following available values:
| InterMineR widget name | InterMineR filter | convertToGeneAnswers categoryType |
|---|---|---|
| go_enrichment_for_gene | biological_process | GO.BP |
| go_enrichment_for_gene | cellular_component | GO.CC |
| go_enrichment_for_gene | molecular_function | GO.MF |
| pathway_enrichment | KEGG pathways data set | KEGG |
| pathway_enrichment | Reactome data set | REACTOME.PATH |
for the categoryType argument of the geneAnswerBuilder function, and can be assigned accordingly in the categoryType argument of the convertToGeneAnswers, therby facilitating the conversion of InterMineR Gene Ontology and Pathway Enrichment analysis results to GeneAnswers-class objects.
# convert to GeneAnswers results for GO terms associated with molecular function
geneanswer_MF_object = convertToGeneAnswers(
enrichmentResult = GO_MF_enrichResult,
geneInput = data.frame(GeneID = as.character(hsa_gene_entrez),
stringsAsFactors = FALSE),
geneInputType = "Gene.primaryIdentifier",
annLib = 'org.Hs.eg.db',
categoryType = "GO.MF"
#enrichCategoryChildName = "Gene.goAnnotation.ontologyTerm.parents.identifier"
)
class(geneanswer_MF_object)## [1] "GeneAnswers"
## attr(,"package")
## [1] "GeneAnswers"summary(geneanswer_MF_object)## This GeneAnswers instance was build from GO.MF based on hyperG test.
## Statistical information of 6 categories with p value less than 0.05 are reported. Other categories are considered as nonsignificant.
## There are 6 categories related to the given 68 genes
##
## Summary of GeneAnswers instance information:
##
## Slot: geneInput
## GeneID
## 1 6833
## 2 1636
## 3 208
## 4 26060
## 5 359
## 6 551
## ......
##
## Slot: testType
## [1] "hyperG"
##
## Slot: pvalueT
## [1] 0.05
##
## Slot: genesInCategory
## $`GO:0005102`
## [1] "1636" "551" "640" "2056" "3077" "3082" "3159" "3172" "3557"
## [10] "3569" "3630" "3643" "3667" "8660" "5468" "5770" "56729" "6934"
## [19] "7422"
##
## $`GO:0005158`
## [1] "3630" "3667" "8660" "5770"
##
## $`GO:0005159`
## [1] "3630" "3643" "3667"
##
## $`GO:0140297`
## [1] "50943" "3159" "3172" "4760" "5468" "6925" "6934" "7466"
##
## $`GO:0001067`
## [1] "50943" "3159" "6927" "6928" "3172" "8462" "4760" "5078" "3651"
## [10] "5325" "5468" "6925" "6934"
##
## $`GO:0044212`
## [1] "50943" "3159" "6927" "6928" "3172" "8462" "4760" "5078" "3651"
## [10] "5325" "5468" "6925" "6934"
##
##
## Slot: geneExprProfile
## NULL
##
## Slot: annLib
## [1] "org.Hs.eg.db"
##
## Slot: categoryType
## [1] "GO.MF"
##
## Slot: enrichmentInfo
## genes in Category percent in the observed List
## GO:0005102 19 0.27941176
## GO:0005158 4 0.05882353
## GO:0005159 3 0.04411765
## GO:0140297 8 0.11764706
## GO:0001067 13 0.19117647
## GO:0044212 13 0.19117647
## percent in the genome fold of overrepresents odds ratio
## GO:0005102 0.0936250151 2.984371 3.753821
## GO:0005158 0.0012725730 46.224090 49.050595
## GO:0005159 0.0009695794 45.501838 47.555769
## GO:0140297 0.0203611683 5.778011 6.415079
## GO:0001067 0.0552054296 3.463001 4.045165
## GO:0044212 0.0550842322 3.470621 4.054585
## p value fdr p value
## GO:0005102 0.003753491 0.003753491
## GO:0005158 0.004425289 0.004425289
## GO:0005159 0.023859182 0.023859182
## GO:0140297 0.023910795 0.023910795
## GO:0001067 0.025815502 0.025815502
## GO:0044212 0.033633007 0.033633007#slotNames(geneanswer_MF_object)Briefly, in the following sections we present how to use several functions of the GeneAnswers package to visualize the results of the Gene Ontology Enrichment analysis on the PL_DiabetesGenes gene list.
For more information about the usage of GeneAnswers package, the user should look at the respective documentation.
3.2 PieChart and BarPlot graphs
# Make GeneAnswers Instance readable
geneanswer_object_readable = geneAnswersReadable(geneanswer_object)# GeneAnswers pieChart
geneAnswersChartPlots(geneanswer_object_readable,
chartType='pieChart',
sortBy = 'correctedPvalue',
top = 3)
Figure 1: Pie chart of the top three GO terms with the smallest FDR-corrected p-values.
# GeneAnswers barplot
geneAnswersChartPlots(geneanswer_object_readable,
chartType='barPlot',
sortBy = 'correctedPvalue',
top = 5)
Figure 2: Bar plot of the top five GO terms with the smallest FDR-corrected p-values.
3.3 Concept-gene network
# generate concept-gene network
geneAnswersConceptNet(geneanswer_object,
colorValueColumn=NULL,
centroidSize='correctedPvalue',
output='interactive',
geneSymbol = TRUE,
catTerm = TRUE,
catID = TRUE,
showCats = 1:5)
Figure 3: Screen shot of concept-gene network for the first five GO terms. The size of nodes is proportional to the number of genes in each GO term.
The color of gene nodes can be changed based on the numeric values associated with the genes.
# generate concept-gene network
# for visualization purposes add a column of RANDOM numbers in geneInput
set.seed(123)
geneanswer_object@geneInput$random_values = rnorm(nrow(geneanswer_object@geneInput))
geneAnswersConceptNet(geneanswer_object,
colorValueColumn=2,
centroidSize='correctedPvalue',
output='interactive',
geneSymbol = TRUE,
catTerm = TRUE,
catID = TRUE,
showCats = 1:5)
Figure 4: Screen shot of concept-gene network for the first five GO terms. Random numeric values have been assigned as column to geneInput in order to visualize the ability to add color to gene nodes. The size of nodes is proportional to the number of genes in each GO term.
3.4 Gene Ontology structure network
# visualize the relations between the first two GO terms
geneAnswersConceptRelation(geneanswer_object,
directed=TRUE,
netMode='connection',
catTerm=TRUE,
catID=TRUE,
showCats = 1:2)
Figure 5: Screen shot of Gene Ontology structure network for the first two GO terms. The size of nodes is proportional to number of genes in these GO terms. The color of nodes stand for how relative the given genes are to the GO terms. More red, more relative. The given GO terms are yellow framed dots with dark purple edges as connections.
3.5 Gene interaction network
The functions of GeneAnswers package make use of several Bioconductor Annotation Packages, which are primarily based on Entrez Gene identifiers.
In our example, GeneAnswers uses the annotation package ‘org.Hs.eg.db’, which was specified previously in convertToGeneAnswers function, to retrieve gene interactions for the specified genes.
# visualize interactions between the first five genes
buildNet(getGeneInput(geneanswer_object)[1:5,1],
idType='GeneInteraction',
layers=2,
filterGraphIDs=getGeneInput(geneanswer_object)[,1],
filterLayer=2,
netMode='connection')
Figure 6: Screen shot of gene interaction network for the first five genes of PL_DiabetesGenes. The large black dots with yellow frame stand for the five given genes. They also connect to other genes by dark-blue-purple edges. Small black dots represent the other genes from getGeneInput. Small white dots are genes that are not in the genes from getGeneInput, but interact with these genes.
3.6 Gene Ontology terms - Genes cross tabulation
# generate GO terms - Genes cross tabulation
geneAnswersHeatmap(geneanswer_object,
catTerm=TRUE,
geneSymbol=TRUE,
showCats = 5:10,
cex.axis = 0.75)## initial value 29.026202
## final value 29.026202
## converged
## initial value 1.171499
## iter 5 value 0.129679
## iter 10 value 0.010251
## iter 10 value 0.007274
## iter 10 value 0.007075
## final value 0.007075
## converged
4 Visualization of InterMineR Publication Enrichment analysis results
4.1 Perform Publication Enrichment analysis
Although GeneAnswers package is designed for performing enrichment analysis on genes and reporting enriched pathways, it is possible to utilize some of its functions to visualize the results of the various enrichment widgets used by InterMineR.
Specifically in the current example we will show how to perform visualizations on publication enrichment analysis results.
# publication enrichment analysis
publication_enrichResult = doEnrichment(
im = im.human,
ids = hsa_gene_entrez,
widget = "publication_enrichment")Since we are performing conversion of enrichment results that are not related to pathways there are three things to notice here:
- We are deliberately assign Gene.symbol instead of Gene.primaryIdentifier values as geneInput. This will result in a more readable concept-gene network and can be changed in order to generate the gene interaction network.
- annLib argument is assigned with ‘org.Hs.eg.db’ as before
- categoryType is assigned with NULL since we no longer use GO terms but PubMed IDs (Gene.publications.pubMedId) as annotation categories.
# convert to GeneAnswer-class
# use Gene.symbol values instead of Gene.primaryIdentifier (ENTREZ)
geneanswer_pubs_object = convertToGeneAnswers(
enrichmentResult = publication_enrichResult,
geneInput = data.frame(GeneID = as.character(hsa_gene_names),
stringsAsFactors = FALSE),
geneInputType = "Gene.symbol",
annLib = 'org.Hs.eg.db',
categoryType = NULL,
enrichCategoryChildName = "Gene.publications.pubMedId"
)
class(geneanswer_pubs_object)## [1] "GeneAnswers"
## attr(,"package")
## [1] "GeneAnswers"summary(geneanswer_pubs_object)## This GeneAnswers instance was build from based on hyperG test.
## Statistical information of 1319 categories with p value less than 0.05 are reported. Other categories are considered as nonsignificant.
## There are 1319 categories related to the given 68 genes
##
## Summary of GeneAnswers instance information:
##
## Slot: geneInput
## GeneID
## 1 ABCC8
## 2 ACE
## 3 AKT2
## 4 APPL1
## 5 AQP2
## 6 AVP
## ......
##
## Slot: testType
## [1] "hyperG"
##
## Slot: pvalueT
## [1] 0.05
##
## Slot: genesInCategory
## $`29792621`
## [1] "ABCC8" "APPL1" "BLK" "CEL" "GCK" "HNF1A" "HNF1B"
## [8] "HNF4A" "INS" "KCNJ11" "KLF11" "NEUROD1" "PAX4" "PDX1"
##
## $`19578796`
## [1] "ABCC8" "ACE" "CAPN10" "CCR5" "ENPP1" "GCGR" "GCK" "HNF1A"
## [9] "HNF4A" "IL6" "INS" "INSR" "IRS1" "KCNJ11" "LIPC" "PAX4"
## [17] "PDX1" "PON1" "PPARG" "PTPN1" "RETN" "VEGFA"
##
## $`19913121`
## [1] "ABCC8" "ACE" "AKT2" "AVPR2" "CCR5" "CDKAL1" "CEL"
## [8] "CTLA4" "EIF2AK3" "ENPP1" "EPO" "GCK" "HFE" "HGF"
## [15] "HMGA1" "HNF1A" "HNF1B" "HNF4A" "IGF2BP2" "IL1RN" "IL2RA"
## [22] "IL6" "INS" "INSR" "IRS1" "IRS2" "ITPR3" "KCNJ11"
## [29] "LIPC" "NEUROD1" "PDX1" "PON1" "PPARG" "PTPN22" "RETN"
## [36] "SLC2A2" "SOD2" "TCF7L2" "VEGFA" "WFS1"
##
## $`20628086`
## [1] "ABCC8" "ACE" "AKT2" "AVPR2" "CCR5" "CDKAL1" "CEL"
## [8] "CTLA4" "EIF2AK3" "ENPP1" "EPO" "GCK" "HFE" "HGF"
## [15] "HMGA1" "HNF1A" "HNF1B" "HNF4A" "IGF2BP2" "IL1RN" "IL2RA"
## [22] "IL6" "INS" "INSR" "IRS1" "IRS2" "ITPR3" "KCNJ11"
## [29] "LIPC" "NEUROD1" "PDX1" "PON1" "PPARG" "PTPN22" "RETN"
## [36] "SLC2A2" "SOD2" "TCF7L2" "VEGFA" "WFS1"
##
## $`20682687`
## [1] "ABCC8" "ACE" "CAPN10" "GCK" "HNF1A" "HNF1B" "HNF4A"
## [8] "IL6" "IRS1" "KCNJ11" "LIPC" "NEUROD1" "PDX1" "PPARG"
## [15] "PTPN1"
##
## $`20384434`
## [1] "ABCC8" "CDKAL1" "HNF1B" "HNF4A" "IGF2BP2" "IL6" "IRS1"
## [8] "KCNJ11" "LIPC" "PPARG" "SLC30A8" "TCF7L2" "WFS1"
##
## ......
##
## Slot: geneExprProfile
## NULL
##
## Slot: annLib
## [1] "org.Hs.eg.db"
##
## Slot: categoryType
## NULL
##
## Slot: enrichmentInfo
## genes in Category percent in the observed List
## 29792621 14 0.2058824
## 19578796 22 0.3235294
## 19913121 40 0.5882353
## 20628086 40 0.5882353
## 20682687 15 0.2205882
## 20384434 13 0.1911765
## percent in the genome fold of overrepresents odds ratio
## 29792621 0.0003684986 558.70588 703.29630
## 19578796 0.0040008423 80.86533 119.06178
## 19913121 0.0584333544 10.06677 23.01931
## 20628086 0.0580122131 10.13985 23.19679
## 20682687 0.0013423879 164.32526 210.54939
## 20384434 0.0007106759 269.00654 332.35354
## p value fdr p value
## 29792621 4.099445e-34 4.099445e-34
## 19578796 2.636278e-31 2.636278e-31
## 19913121 4.500934e-27 4.500934e-27
## 20628086 4.534921e-27 4.534921e-27
## 20682687 3.483531e-25 3.483531e-25
## 20384434 9.462024e-25 9.462024e-25
## ......#slotNames(geneanswer_pubs_object)4.2 PieChart and BarPlot graphs
# GeneAnswers pieChart
geneAnswersChartPlots(geneanswer_pubs_object,
chartType='pieChart',
sortBy = 'correctedPvalue',
top = 5)
Figure 7: Pie chart of the top five PubMed IDs with the smallest FDR-corrected p-values.
# GeneAnswers barplot
geneAnswersChartPlots(geneanswer_pubs_object,
chartType='barPlot',
sortBy = 'correctedPvalue',
top = 5)
Figure 8: Bar plot of the top five PubMed IDs with the smallest FDR-corrected p-values.
4.3 Concept-gene network
It is important to note that the arguments catTerm and catID are assigned with FALSE. This is because the function is not expecting PubMed IDs as annotation categories and it will fail to convert them to short descriptions (as it would do with GO terms or KEGG and REACTOME pathways).
# generate concept-gene network
geneAnswersConceptNet(geneanswer_pubs_object,
colorValueColumn=NULL,
centroidSize='correctedPvalue',
output='interactive',
catTerm = FALSE,
catID = FALSE,
showCats = 1:5)
Figure 9: Screen shot of concept-gene network for the first five PubMed IDs. The size of nodes is proportional to the number of genes in each PubMed ID.
4.4 Gene interaction network
To generate a gene interaction network it is necessary to replace Gene.symbol with Gene.primaryIdentifier values because buildNet function retrieves gene interactions from the ‘org.Hs.eg.db’ package, which uses ENTREZ identifiers for mapping.
# copy the existing GeneAnswer-class object
geneanswer_pubs_object2 = geneanswer_pubs_object
# replace Gene.symbol with Gene.primaryIdentifier values
geneanswer_pubs_object2@geneInput = data.frame(
GeneID = as.character(hsa_gene_entrez),
stringsAsFactors = FALSE)
# generate gene interaction network
buildNet(getGeneInput(geneanswer_pubs_object2)[1:5,1],
idType='GeneInteraction',
layers=2,
filterGraphIDs=getGeneInput(geneanswer_pubs_object)[,1],
filterLayer=2,
netMode='connection')
Figure 10: Screen shot of gene interaction network for the first five genes of PL_DiabetesGenes. The large black dots with yellow frame stand for the five given genes. They also connect to other genes by dark-blue-purple edges. Small black dots represent the other genes from getGeneInput. Small white dots are genes that are not in the genes from getGeneInput, but interact with these genes.
5 System info
sessionInfo()## R version 3.6.1 (2019-07-05)
## Platform: x86_64-pc-linux-gnu (64-bit)
## Running under: Ubuntu 18.04.2 LTS
##
## Matrix products: default
## BLAS: /home/biocbuild/bbs-3.9-bioc/R/lib/libRblas.so
## LAPACK: /home/biocbuild/bbs-3.9-bioc/R/lib/libRlapack.so
##
## locale:
## [1] LC_CTYPE=en_US.UTF-8 LC_NUMERIC=C
## [3] LC_TIME=C LC_COLLATE=C
## [5] LC_MONETARY=en_US.UTF-8 LC_MESSAGES=en_US.UTF-8
## [7] LC_PAPER=en_US.UTF-8 LC_NAME=C
## [9] LC_ADDRESS=C LC_TELEPHONE=C
## [11] LC_MEASUREMENT=en_US.UTF-8 LC_IDENTIFICATION=C
##
## attached base packages:
## [1] parallel stats4 stats graphics grDevices utils datasets
## [8] methods base
##
## other attached packages:
## [1] org.Hs.eg.db_3.8.2 GO.db_3.8.2 GeneAnswers_2.26.0
## [4] RColorBrewer_1.1-2 Heatplus_2.30.0 MASS_7.3-51.4
## [7] RSQLite_2.1.1 annotate_1.62.0 XML_3.98-1.20
## [10] AnnotationDbi_1.46.0 IRanges_2.18.1 S4Vectors_0.22.0
## [13] Biobase_2.44.0 BiocGenerics_0.30.0 RCurl_1.95-4.12
## [16] bitops_1.0-6 igraph_1.2.4.1 InterMineR_1.6.1
## [19] BiocStyle_2.12.0
##
## loaded via a namespace (and not attached):
## [1] Rcpp_1.0.1 lattice_0.20-38
## [3] Biostrings_2.52.0 digest_0.6.20
## [5] R6_2.4.0 GenomeInfoDb_1.20.0
## [7] chron_2.3-53 evaluate_0.14
## [9] httr_1.4.0 sqldf_0.4-11
## [11] zlibbioc_1.30.0 curl_3.3
## [13] blob_1.1.1 Matrix_1.2-17
## [15] rmarkdown_1.13 gsubfn_0.7
## [17] proto_1.0.0 BiocParallel_1.18.0
## [19] downloader_0.4 stringr_1.4.0
## [21] bit_1.1-14 DelayedArray_0.10.0
## [23] compiler_3.6.1 xfun_0.8
## [25] pkgconfig_2.0.2 htmltools_0.3.6
## [27] SummarizedExperiment_1.14.0 GenomeInfoDbData_1.2.1
## [29] bookdown_0.11 matrixStats_0.54.0
## [31] grid_3.6.1 RBGL_1.60.0
## [33] jsonlite_1.6 xtable_1.8-4
## [35] DBI_1.0.0 magrittr_1.5
## [37] graph_1.62.0 stringi_1.4.3
## [39] XVector_0.24.0 xml2_1.2.0
## [41] tools_3.6.1 RJSONIO_1.3-1.2
## [43] bit64_0.9-7 yaml_2.2.0
## [45] BiocManager_1.30.4 GenomicRanges_1.36.0
## [47] memoise_1.1.0 knitr_1.23References
Feng, Gang, Pan Du, Nancy L Krett, Michael Tessel, Steven Rosen, Warren A Kibbe, and Simon M Lin. 2010. “A collection of bioconductor methods to visualize gene-list annotations.” BMC Research Notes 3 (1):10. https://doi.org/10.1186/1756-0500-3-10.
Feng, Gang, Pamela Shaw, Steven T. Rosen, Simon M. Lin, and Warren A. Kibbe. 2012. “Using the Bioconductor GeneAnswers Package to Interpret Gene Lists.” In Methods in Molecular Biology (Clifton, N.j.), 802:101–12. https://doi.org/10.1007/978-1-61779-400-1_7.
Huang, Lei, Gang Feng, Pan Du, Tian Xia, Xishu Wang, Jing Wen, Warren Kibbe, and Simon Lin. 2014. “GeneAnswers: Integrated Interpretation of Genes.” Bioconductor. https://www.bioconductor.org/packages/release/bioc/html/GeneAnswers.html.
Kalderimis, Alex, Rachel Lyne, Daniela Butano, Sergio Contrino, Mike Lyne, Joshua Heimbach, Fengyuan Hu, et al. 2014. “InterMine: extensive web services for modern biology.” Nucleic Acids Research 42 (Web Server issue). Oxford University Press:W468–72. https://doi.org/10.1093/nar/gku301.
Smith, R. N., J. Aleksic, D. Butano, A. Carr, S. Contrino, F. Hu, M. Lyne, et al. 2012. “InterMine: a flexible data warehouse system for the integration and analysis of heterogeneous biological data.” Bioinformatics 28 (23):3163–5. https://doi.org/10.1093/bioinformatics/bts577.