Generalised gene set testing for Illumina's methylation array data

Share:

Description

Given a user specified list of gene sets to test, gsameth tests whether significantly differentially methylated CpG sites are enriched in these gene sets.

Usage

1
gsameth(sig.cpg, all.cpg = NULL, collection, array.type=c("450K","EPIC"),plot.bias = FALSE, prior.prob = TRUE)

Arguments

sig.cpg

character vector of significant CpG sites to test for gene set enrichment

all.cpg

character vector of all CpG sites tested. Defaults to all CpG sites on the array.

collection

a list of user specified gene sets to test. Can also be a single character vector gene set. Gene identifiers must be Entrez Gene IDs.

array.type

the Illumina methylation array used. Options are "450K" or "EPIC". Defaults to "450K".

plot.bias

logical, if true a plot showing the bias due to the differing numbers of probes per gene will be displayed

prior.prob

logical, if true will take into account the probability of significant differentially methylation due to numbers of probes per gene. If false, a hypergeometric test is performed ignoring any bias in the data.

Details

This function extends gometh, which only tests GO and KEGG pathways. gsameth can take a list of user specified gene sets and test whether the significant CpG sites are enriched in these pathways. gsameth maps the CpG sites to Entrez Gene IDs and tests for pathway enrichment using a hypergeometric test, taking into account the number of CpG sites per gene on the 450K/EPIC arrays. Please note the gene ids for the collection of gene sets must be Entrez Gene IDs.

Geeleher et al. (2013) showed that a severe bias exists when performing gene set analysis for genome-wide methylation data that occurs due to the differing numbers of CpG sites profiled for each gene. gsameth and gometh is based on the goseq method (Young et al., 2010). If prior.prob is set to FALSE, then prior probabilities are not used and it is assumed that each gene is equally likely to have a significant CpG site associated with it.

Genes associated with each CpG site are obtained from the annotation package IlluminaHumanMethylation450kanno.ilmn12.hg19 if the array type is "450K". For the EPIC array, the annotation package IlluminaHumanMethylationEPICanno.ilm10b2.hg19 is used. In order to get a list which contains the mapped Entrez gene IDS, please use the getMappedEntrezIDs function.

Value

A data frame with a row for each gene set and the following columns:

N

number of genes in the gene set

DE

number of genes that are differentially methylated

P.DE

p-value for over-representation of the gene set

FDR

False discovery rate, calculated using the method of Benjamini and Hochberg (1995).

Author(s)

Belinda Phipson

References

Phipson, B., Maksimovic, J., and Oshlack, A. (2016). missMethyl: an R package for analysing methylation data from Illuminas HumanMethylation450 platform. Bioinformatics, 15;32(2), 286–8.

Geeleher, P., Hartnett, L., Egan, L. J., Golden, A., Ali, R. A. R., and Seoighe, C. (2013). Gene-set analysis is severely biased when applied to genome-wide methylation data. Bioinformatics, 29(15), 1851–1857.

Young, M. D., Wakefield, M. J., Smyth, G. K., and Oshlack, A. (2010). Gene ontology analysis for RNA-seq: accounting for selection bias. Genome Biology, 11, R14.

Ritchie, M. E., Phipson, B., Wu, D., Hu, Y., Law, C. W., Shi, W., and Smyth, G. K. (2015). limma powers differential expression analyses for RNA-sequencing and microarray studies. Nucleic Acids Research, gkv007.

Benjamini, Y., and Hochberg, Y. (1995). Controlling the false discovery rate: a practical and powerful approach to multiple testing. Journal of the Royal Statistical Society Series, B, 57, 289-300.

See Also

gometh,getMappedEntrezIDs

Examples

 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
library(IlluminaHumanMethylation450kanno.ilmn12.hg19)
library(org.Hs.eg.db)
library(limma)
ann <- getAnnotation(IlluminaHumanMethylation450kanno.ilmn12.hg19)

# Randomly select 1000 CpGs to be significantly differentially methylated
sigcpgs <- sample(rownames(ann),1000,replace=FALSE)

# All CpG sites tested
allcpgs <- rownames(ann)

# Use org.Hs.eg.db to extract a GO term
GOtoID <- toTable(org.Hs.egGO2EG)
setname1 <- GOtoID$go_id[1]
setname1
keep.set1 <- GOtoID$go_id %in% setname1
set1 <- GOtoID$gene_id[keep.set1]
setname2 <- GOtoID$go_id[2]
setname2
keep.set2 <- GOtoID$go_id %in% setname2
set2 <- GOtoID$gene_id[keep.set2]

# Make the gene sets into a list
sets <- list(set1, set2)
names(sets) <- c(setname1,setname2)

# Testing with prior probabilities taken into account
# Plot of bias due to differing numbers of CpG sites per gene
gst <- gsameth(sig.cpg = sigcpgs, all.cpg = allcpgs, collection = sets, plot.bias = TRUE, prior.prob = TRUE)
topGSA(gst)

# Testing ignoring bias
gst.bias <- gsameth(sig.cpg = sigcpgs, all.cpg = allcpgs, collection = sets, prior.prob = FALSE)
topGSA(gst.bias)