COHCAP.site: CpG Site Differential Methylation Analysis

In cwarden45/COHCAP: CpG Island Analysis Pipeline for Illumina Methylation Array and Targeted BS-Seq Data

CpG Site Differential Methylation Analysis

Description

Provides statistics for CpG sites as well as a list of differentially methylated sites. Can also provide .wig files for visualization in IGV, UCSC Genome Browser, etc.

List of differentially methylated sites and .wig files will be created in the "CpG_Site" folder. Table of statistics for all CpG sites will be created in the "Raw_Data" folder.

Usage

COHCAP.site(sample.file, beta.table, project.name, project.folder,
				methyl.cutoff=0.7, unmethyl.cutoff = 0.3, paired=FALSE,
				delta.beta.cutoff = 0.2, pvalue.cutoff=0.05,
				fdr.cutoff=0.05, ref="none", num.groups=2,
				lower.cont.quantile=0, upper.cont.quantile=1,
				create.wig = "avg", alt.pvalue="none",
				plot.heatmap=TRUE, output.format = "txt", heatmap.dist.fun="Euclidian")

Arguments

sample.file	Tab-delimited text file providing group attributions for all samples considered for analysis.
beta.table	Data frame with CpG sites in columns (with DNA methylation represented as beta values or percentage methylation), samples in columns, and CpG site annotations are included (in columns 2-5). The COHCAP.annotate function automatically creates this file.
project.name	Name for COHCAP project. This determines the names for output files.
project.folder	Folder for COHCAP output files
methyl.cutoff	Minimum beta or percentage methylation value to be used to define a methylated CpG site. Default is 0.7 (used for beta values), which would correspond to 70 percent methylation. Used for either 1-group or 2-group comparison.
unmethyl.cutoff	Minimum beta or percentage methylation value to be used to define an unmethylated CpG site. Default is 0.3 (used for beta values), which would correspond to 30 percent methylation. Used for either 1-group or 2-group comparison.
delta.beta.cutoff	The minimum absolute value for delta-beta values (mean treatment beta - mean reference beta) to define a differentially methylated CpG site. Only used for 2-group comparison.
pvalue.cutoff	Maximum p-value allowed to define a site as differentially methylated. Used only for comparisons with at least 2 groups (with 3 replicates per group)
fdr.cutoff	Maximum False Discovery Rate (FDR) allowed to define a site as differentially methylated. Used only for comparisons with at least 2 groups (with 3 replicates per group)
ref	Reference group used to define baseline methylation levels. Set to "continuous" for a continuous primary variable. Otherwise, only used for 2-group comparison.
num.groups	Number of groups described in sample description file. COHCAP algorithm differs when analysing 1-group, 2-group, or >2-group comparisons. Not used if "ref" is set to "continuous" (for linear-regression of a continuous variable).
create.wig	Set to "avg" to create average beta (per group) and delta-beta values. Set to "sample" to create .wig files for each sample. Set to "avg.and.sample" to create average, delta-beta, and per-sample .wig files. Seto to "none" to avoid creating .wig files In the standalone version of COHCAP, this was only an option when using the "Average by Site" workflow (because that was the only situation where the analysis method matched the visualization). .wig files are defined with respect to hg19 (for pre-defined annotation files) and can be visualized using IGV, UCSC Genome Browser, etc.
plot.heatmap	Logical value: Should heatmap be created to visualize CpG site differential methylation? For best comparison to island heatmap, please use same parameters at site and island level.
lower.cont.quantile	For continuous analysis, what beta quantile should be the lower threshold for calculating delta-beta values? Default = 0 (minimum)
upper.cont.quantile	For continuous analysis, what beta quantile should be the upper threshold for calculating delta-beta values? Default = 1 (maximum)
alt.pvalue	Use alternative strategies for p-value calculations. Be careful that the workflow matches the p-value calculation. For 'rANOVA.1way', use ANOVA (R function) instead of t-test (for 1-variable, 2-group comparison). Might be helpful when SD is 0. For 'cppANOVA.1way', use ANOVA (C++ code) instead of t-test (for 1-variable, 2-group comparison). Helps decrease run-time relative to R-code. For 'cppANOVA.2way', use C++ code instead of R function for t-test (for 2-variable, 2-group comparison). Helps decrease run-time relative to R-code, but p-value may be different. For this implementation, I require having at replicates for each interaction term (such as having replicate treatmentswith multiple backgrounds, cell lines, etc.). If each sample has exactly one pair ( as may be the case with tumor-normal pairs), and you need to decrease the COHCAP run-time, you may consider using 'cppPairedTtest'. For 'cppWelshTtest', use C++ code instead of R function for t-test (for 1-variable, 2-group comparison). Helps decrease run-time relative to R-code, but p-value will be different than t.test() function (C++ code assumes unequal variance between groups). For 'cppPairedTtest', use C++ code instead of R function for t-test (for 2-variable, 2-group comparison). Helps decrease run-time relative to R-code. T-test not usually paired in COHCAP, so p-value will be different than for 1-variable test. May be useful if you need to speed up code and all measurements are paired. For 'cppLmResidual.1var', use C++ code instead of R function for linear regression (t-test for residuals). Only valid for continuous analysis with 1 variable. WARNING: This code may less sensitive than normal lm() or ANOVA with smaller sample sizes (such as n=6). For 'RcppArmadillo.fastLmPure', use 'fastLmPure' function within RcppArmadillo for linear regression, with R pt() t-distribution for p-value calculation. This can help decrease the run-time, relative to the lm() function that is used by default. Can be 'none','rANOVA.1way','RcppArmadillo.fastLmPure', 'cppANOVA.1way', 'cppANOVA.2way','cppWelshTtest', or 'cppPairedTtest'.
heatmap.dist.fun	Distance metric for clustering in heatmap. Can be 'Euclidian' or 'Pearson Dissimilarity'.
paired	A logical value: Is there any special pairing between samples in different groups? If so, the pairing variable must be specified in the 3rd column of the sample description file. Used for p-value calculation, so this only applies to comparisons with at least 2 groups. If you have a secondary continuous variable (like age), you can set paired to "continuous". COHCAP will then perform linear-regression analysis (converting primary categorical variable into continuous variable, if necessary)
output.format	Format for output tables: 'xls' for Excel file, 'csv' for comma-separated file, or 'txt' for tab-delimited text file.

Value

Data frame of average beta (or percentage methylation) statistics and/or p-value / false discovery rate statistics.

The content of the data frame depends upon the number of groups specified for analysis (avg.beta only for 1-group; avg.beta, delta.beta, p-value, and FDR for 2-group; p-value and FDR only for >2 groups).

This data frame is used for CpG island analysis.

See Also

COHCAP Discussion Group: http://sourceforge.net/p/cohcap/discussion/general/

Examples

library("COHCAP")

dir = system.file("extdata", package="COHCAP")
beta.file = file.path(dir,"GSE42308_truncated.txt")
sample.file = file.path(dir,"sample_GSE42308.txt")
project.folder = tempdir()#you may want to use getwd() or specify another folder
project.name = "450k_test"

beta.table = COHCAP.annotate(beta.file, project.name, project.folder,
				platform="450k-UCSC")
filtered.sites = COHCAP.site(sample.file, beta.table, project.name,
				project.folder, ref="parental")

cwarden45/COHCAP documentation built on April 16, 2024, 3:17 a.m.