DMRcaller (version 1.4.2)

computeDMRs: Compute DMRs

Description

This function computes the differentially methylated regions between two conditions.

Usage

computeDMRs(methylationData1, methylationData2, regions = NULL, context = "CG", method = "noise_filter", windowSize = 100, kernelFunction = "triangular", lambda = 0.5, binSize = 100, test = "fisher", pValueThreshold = 0.01, minCytosinesCount = 4, minProportionDifference = 0.4, minGap = 200, minSize = 50, minReadsPerCytosine = 4, cores = 1)

Arguments

methylationData1

the methylation data in condition 1 (see methylationDataList).

methylationData2

the methylation data in condition 2 (see methylationDataList).

regions

a GRanges object with the regions where to compute the DMRs. If NULL, the DMRs are computed genome-wide.

context

the context in which the DMRs are computed ("CG", "CHG" or "CHH").

method

the method used to compute the DMRs ("noise_filter", "neighbourhood" or "bins"). The "noise_filter" method uses a triangular kernel to smooth the number of reads and then performs a statistical test to determine which regions dispay different levels of methylation in the two conditions. The "neighbourhood" method computates differentially methylated cytosines. Finally, the "bins" method partiones the genome into equal sized tilling bins and performs the statistical test between the two conditions in each bin. For all three methods, the cytosines or bins are then merged into DMRs without affecting the inital parameters used when calling the differentiall methylated cytosines/bins (p-value, difference in methylation levels, minimum number of reads per cytosine).

windowSize

the size of the triangle base measured in nucleotides. This parameter is required only if the selected method is "noise_filter".

kernelFunction

a character indicating which kernel function to be used. Can be one of "uniform", "triangular", "gaussian" or "epanechnicov". This is required only if the selected method is "noise_filter".

lambda

numeric value required for the Gaussian filter (K(x) = exp(-lambda*x^2)). This is required only if the selected method is "noise_filter" and the selected kernel function is "gaussian".

binSize

the size of the tiling bins in nucleotides. This parameter is required only if the selected method is "bins".

test

the statistical test used to call DMRs ("fisher" for Fisher's exact test or "score" for Score test).

pValueThreshold

DMRs with p-values (when performing the statistical test; see test) higher or equal than pValueThreshold are discarded. Note that we adjust the p-values using the Benjamini and Hochberg's method to control the false discovery rate.

minCytosinesCount

DMRs with less cytosines in the specified context than minCytosinesCount will be discarded.

minProportionDifference

DMRs where the difference in methylation proportion between the two conditions is lower than minProportionDifference are discarded.

minGap

DMRs separated by a gap of at least minGap are not merged. Note that only DMRs where the change in methylation is in the same direction are joined.

minSize

DMRs with a size smaller than minSize are discarded.

minReadsPerCytosine

DMRs with the average number of reads lower than minReadsPerCytosine are discarded.

cores

the number of cores used to compute the DMRs.

Value

the DMRs stored as a GRanges object with the following metadata columns:

direction

a number indicating whether the region lost (-1) or gain (+1) methylation in condition 2 compared to condition 1.

context

the context in which the DMRs was computed ("CG", "CHG" or "CHH").

sumReadsM1

the number of methylated reads in condition 1.

sumReadsN1

the total number of reads in condition 1.

proportion1

the proportion methylated reads in condition 1.

sumReadsM2

the number of methylated reads in condition 2.

sumReadsN2

the total number reads in condition 2.

proportion2

the proportion methylated reads in condition 2.

cytosinesCount

the number of cytosines in the DMR.

regionType

a string indicating whether the region lost ("loss") or gained ("gain") methylation in condition 2 compared to condition 1.

pValue

the p-value (adjusted to control the false discovery rate with the Benjamini and Hochberg's method) of the statistical test when the DMR was called.

See Also

filterDMRs, mergeDMRsIteratively, analyseReadsInsideRegionsForCondition and DMRsNoiseFilterCG

Examples

# load the methylation data
data(methylationDataList)

# the regions where to compute the DMRs
regions <- GRanges(seqnames = Rle("Chr3"), ranges = IRanges(1,1E5))

# compute the DMRs in CG context with noise_filter method
DMRsNoiseFilterCG <- computeDMRs(methylationDataList[["WT"]],
                     methylationDataList[["met1-3"]], regions = regions,
                     context = "CG", method = "noise_filter",
                     windowSize = 100, kernelFunction = "triangular",
                     test = "score", pValueThreshold = 0.01,
                     minCytosinesCount = 4, minProportionDifference = 0.4,
                     minGap = 200, minSize = 50, minReadsPerCytosine = 4,
                     cores = 1)

## Not run: 
# # compute the DMRs in CG context with neighbourhood method
# DMRsNeighbourhoodCG <- computeDMRs(methylationDataList[["WT"]],
#                        methylationDataList[["met1-3"]], regions = regions,
#                        context = "CG", method = "neighbourhood",
#                        test = "score", pValueThreshold = 0.01,
#                        minCytosinesCount = 4, minProportionDifference = 0.4,
#                        minGap = 200, minSize = 50, minReadsPerCytosine = 4,
#                        cores = 1)
# 
# # compute the DMRs in CG context with bins method
# DMRsBinsCG <- computeDMRs(methylationDataList[["WT"]],
#                methylationDataList[["met1-3"]], regions = regions,
#                context = "CG", method = "bins", binSize = 100,
#                test = "score", pValueThreshold = 0.01, minCytosinesCount = 4,
#                minProportionDifference = 0.4, minGap = 200, minSize = 50,
#                minReadsPerCytosine = 4, cores = 1)
# 
# ## End(Not run)