Identifying putative probe-gene pairs
This step is to link enhancer probes with methylation changes to target genes with expression changes and report the putative target gene for selected probes. This is carried out by function get.pair.
For each distal probe with differential methylation the closest 10 upstream genes and the closest 10 downstream genes were tested for correlation between methylation of the probe and expression of the gene. Thus, exactly 20 statistical tests were performed for each probe, as follows. For each probe-gene pair, the samples (all experiment samples and control samples) were divided into two groups: the M group, which consisted of the upper methylation quintile (the 20% of samples with the highest methylation at the enhancer probe), and the U group, which consisted of the lowest methylation quintile (the 20% of samples with the lowest methylation.) For each probe-gene pair tested, the raw p-value Pr was corrected for multiple hypothesis using a permutation approach.
(Yao et al. 2015,Yao, Berman, and Farnham (2015))
| Argument | Description |
|---|---|
| data | A multiAssayExperiment with DNA methylation and Gene Expression data. See createMAE function. |
| nearGenes | Can be either a list containing output of GetNearGenes function or path of rda file containing output of GetNearGenes function. |
| minSubgroupFrac | A number ranging from 0 to 1.0 specifying the percentage of samples used to create groups U (unmethylated) and M (methylated) used to link probes to genes. Default is 0.4 (lowest quintile samples will be in the U group and the highest quintile samples in the M group). |
| permu.size | A number specify the times of permuation. Default is 10000. |
| raw.pvalue | A number specify the raw p-value cutoff for defining signficant pairs. Default is 0.05. It will select the significant P value cutoff before calculating the empirical p-values. |
| Pe | A number specify the empirical p-value cutoff for defining signficant pairs. Default is 0.001. |
| group.col | A column defining the groups of the sample. You can view the available columns using: colnames(MultiAssayExperiment::colData(data)). |
| group1 | A group from group.col. |
| group2 | A group from group.col. |
| mode | A character. Can be “unsupervised” or “supervised”. If unsupervised is set the U (unmethylated) and M (methylated) groups will be selected among all samples based on methylation of each probe. Otherwise U group and M group will set as the samples of group1 or group2 as described below: If diff.dir is “hypo, U will be the group 1 and M the group2. If diff.dir is”hyper" M group will be the group1 and U the group2. |
| diff.dir | A character can be “hypo” or “hyper”, showing differential methylation dirction in group 1. It can be “hypo” which means the probes are hypomethylated in group1; “hyper” which means the probes are hypermethylated in group1; This argument is used only when mode is supervised nad it should be the same value from get.diff.meth function. |
| filter.probes | Should filter probes by selecting only probes that have at least a certain number of samples below and above a certain cut-off. See preAssociationProbeFiltering function. |
| filter.portion | A number specify the cut point to define binary methlation level for probe loci. Default is 0.3. When beta value is above 0.3, the probe is methylated and vice versa. For one probe, the percentage of methylated and unmethylated samples should be above filter.percentage value. Only used if filter.probes is TRUE. See preAssociationProbeFiltering function. |
| filter.percentage | Minimum percentage of samples to be considered in methylated and unmethylated for the filter.portion option. Default 5%. Only used if filter.probes is TRUE. See preAssociationProbeFiltering function. |
# Load results from previous sections
mae <- get(load("mae.rda"))
sig.diff <- read.csv("result/getMethdiff.hypo.probes.significant.csv")
nearGenes <- GetNearGenes(data = mae,
probes = sig.diff$probe,
numFlankingGenes = 20, # 10 upstream and 10 dowstream genes
cores = 1)
Hypo.pair <- get.pair(data = mae,
group.col = "definition",
group1 = "Primary solid Tumor",
group2 = "Solid Tissue Normal",
nearGenes = nearGenes,
minSubgroupFrac = 0.4, # % of samples to use in to create groups U/M
permu.dir = "result/permu",
permu.size = 100, # Please set to 100000 to get significant results
raw.pvalue = 0.05,
Pe = 0.01, # Please set to 0.001 to get significant results
filter.probes = TRUE, # See preAssociationProbeFiltering function
filter.percentage = 0.05,
filter.portion = 0.3,
dir.out = "result",
cores = 1,
label = "hypo")Observation: The distance column in the nearGenes object and in thable getPair.hypo.all.pairs.statistic.csv are the distance to the gene. To update, to the distance to the nearest TSS please use the function addDistNearestTSS. This function was not used default due to time requirements to run for all probes and all their 20 nearest genes, but it is ran for the significant pairs.
Hypo.pair %>% datatable(options = list(scrollX = TRUE))# get.pair automatically save output files.
# getPair.hypo.all.pairs.statistic.csv contains statistics for all the probe-gene pairs.
# getPair.hypo.pairs.significant.csv contains only the significant probes which is
# same with Hypo.pair.
dir(path = "result", pattern = "getPair") ## [1] "getPair.hypo.all.pairs.statistic.csv"
## [2] "getPair.hypo.pairs.significant.csv"
## [3] "getPair.hypo.pairs.statistic.with.empirical.pvalue.csv"Bibliography
Yao, Lijing, Benjamin P Berman, and Peggy J Farnham. 2015. “Demystifying the Secret Mission of Enhancers: Linking Distal Regulatory Elements to Target Genes.” Critical Reviews in Biochemistry and Molecular Biology 50 (6). Taylor & Francis: 550–73.
Yao, Lijing, Hui Shen, Peter W Laird, Peggy J Farnham, and Benjamin P Berman. 2015. “Inferring Regulatory Element Landscapes and Transcription Factor Networks from Cancer Methylomes.” Genome Biology 16 (1). BioMed Central: 105.