preAssociationProbeFiltering: Filtering probes

In ELMER: Inferring Regulatory Element Landscapes and Transcription Factor Networks Using Cancer Methylomes

Description

This function has some filters to the DNA methylation data in each it selects probes to avoid correlations due to non-cancer contamination and for additional stringency.

• Filter 1: We usually call locus unmethylated when the methylation value < 0.3 and methylated when the methylation value > 0.3. Therefore Meth_B is the percentage of methylation value > K. Basically, this step will make sure we have at least a percentage of beta values lesser than K and n percentage of beta values greater K. For example, if percentage is 5%, the number of samples 100 and K = 0.3, this filter will select probes that we have at least 5 (5% of 100%) samples have beta values > 0.3 and at least 5 samples have beta values < 0.3. This filter is importante as true promoters and enhancers usually have a pretty low value (of course purity can screw that up). we often see lots of PMD probes across the genome with intermediate values like 0.4. Choosing a value of 0.3 will certainly give some false negatives, but not compared to the number of false positives we thought we might get without this filter.

Usage

preAssociationProbeFiltering(data, K = 0.3, percentage = 0.05)

Arguments

data	A MultiAssayExperiment with a DNA methylation martrix or a DNA methylation matrix
K	Cut off to consider probes as methylated or unmethylated. Default: 0.3
percentage	The percentage of samples we should have at least considered as methylated and unmethylated

Value

An object with the same class, but with the probes removed.

References

Yao, Lijing, et al. "Inferring regulatory element landscapes and transcription factor networks from cancer methylomes." Genome biology 16.1 (2015): 1. Method section (Linking enhancer probes with methylation changes to target genes with expression changes).

Examples

 random.probe <- runif(100, 0, 1)
 bias_l.probe <- runif(100, 0, 0.3)
 bias_g.probe <- runif(100, 0.3, 1)
 met <- rbind(random.probe,bias_l.probe,bias_g.probe)
 met <- preAssociationProbeFiltering(data = met,  K = 0.3, percentage = 0.05)
 met <- rbind(random.probe,random.probe,random.probe)
 met <- preAssociationProbeFiltering(met,  K = 0.3, percentage = 0.05)
 data <- ELMER:::getdata("elmer.data.example") # Get data from ELMER.data
 data <- preAssociationProbeFiltering(data,  K = 0.3, percentage = 0.05)
 
 cg24741609 <- runif(100, 0, 1)
 cg17468663 <- runif(100, 0, 0.3)
 cg14036402 <- runif(100, 0.3, 1)
 met <- rbind(cg24741609,cg14036402,cg17468663)
 colnames(met) <- paste("sample",1:100)
 exp <- met
 rownames(exp) <- c("ENSG00000141510","ENSG00000171862","ENSG00000171863")
 sample.info <- S4Vectors::DataFrame(primary = paste("sample",1:100),
                                     sample.type = rep(c("Normal", "Tumor"),50))
 rownames(sample.info) <- colnames(exp)
 mae <- createMAE(exp = exp, met = met, colData = sample.info, genome = "hg38") 
 mae <- preAssociationProbeFiltering(mae,  K = 0.3, percentage = 0.05)

ELMER documentation built on Nov. 8, 2020, 4:59 p.m.