dmpClusters: DMP clustering

In genomaths/MethylIT: Methylation Analysis Based on Signal Detection

DMP clustering

Description

Given a 'pDMP' (or 'InfDiv') object carrying DMPs (methylated cytosines) detected in Methyl-IT downstream analysis, function 'dmpClusters' build clusters of DMPs, which can be further tested to identify differentially methylated regions (DMRs) with countTest2 function.

Usage

dmpClusters(
  GR,
  maxDist = 3,
  minNumDMPs = 1,
  maxClustDist = NULL,
  method = c("relaxed", "fixed.int"),
  chromosomes = NULL,
  columns = 9L,
  ignore.strand = TRUE,
  num.cores = detectCores() - 1,
  tasks = 0L,
  verbose = TRUE,
  ...
)

Arguments

GR	An object from 'pDMP' class, which is returned by selectDIMP function.
maxDist	maximum distance at which two reported bases sites from the same cluster can be separated. Default: maxDist = 3.
minNumDMPs	Minimum number of DMPs inside of each cluster. Default: minNumDMPs = 1.
maxClustDist	Clusters separated by a distance lesser than maxClustDist' positions are merged. Default: maxClustDist = NULL. If 0 < maxClustDist < maxDist or maxClustDist = NULL, then maxClustDist will be recalculated as maxClustDist = maxDist + 1.
method	Two different approaches are implemented to clustering DMPs: "relaxed": DMP ranges which are separated by a distance less than 'maxClustDist' are merged and ranges with less than 'minNumDMPs' are removed. "fixed.int": It will generate a partition where the distance between consecutive DMPs is not greater than 'maxDist'. Ranges with less than 'minNumDMPs' are removed. If, additionally, a value maxClustDist > 0 is provided, then the "relaxed" approach is applied to the ranges from the first step.
chromosomes	vector of characters labeling the chromosomes included in the analysis. Default: chromosomes = NULL (all chromosomes are included).
columns	An integer number corresponding to the specific column(s) to use from the meta-column of each GRanges. Default values is 9 (the column carrying to the Hellinger divergence values).
ignore.strand	Same as in findOverlaps-methods.
num.cores, tasks	integer(1). The number of cores to use, i.e. at most how many child processes will be run simultaneously (see bplapply function from BiocParallel package).The number of tasks per job. value must be a scalar integer >= 0L (see MulticoreParam from BiocParallel package).
verbose	if TRUE, prints the function log to stdout.
...	Further parameters for uniqueGRanges function.

Details

Two algorithmic approaches are implemented, named: "relaxed" and "fixed.int" (see the description of parameter 'method'). The "fixed.int" is mostly addressed to find specific methylation patterns, but the price is the number of DMRs found is lower.

The number of DMPs reported in each cluster corresponds to the numbers of sites inside the cluster where DMPs were found in at least one of the samples (from control or from treatment). That is, dmpClusters is only a tool to locate regions with high density of DMPs from all the samples. It does not detect DMRs. It is assumed that only DMP coordinates are given in the 'GR' object. That is, all the sites provided are considered in the computation.

Value

A GRanges object with the numbers of positions inside each cluster, where DMPs were reported in at least one of the samples.

Author(s)

Robersy Sanchez (https://genomaths.com).

References

1. Sanchez R, Mackenzie SA (2016) Information Thermodynamics of Cytosine DNA Methylation. PLOS ONE 11(3): e0150427. https://doi.org/10.1371/journal.pone.0150427

Examples

## Get a dataset of dmps from the package
data(dmps)

## Build clusters of DMPs taking into account the DNA strand
x1 = dmpClusters(GR = dmps, maxDist = 7, minNumDMPs = 6,
                 method = "fixed.int", ignore.strand = FALSE,
                 verbose = FALSE)
data.frame(x1)

## Not run: 
## Build clusters of DMPs ignoring DNA strand and maxClustDist = 7
x2 = dmpClusters(GR = dmps, maxDist = 7, minNumDMPs = 6,
                  maxClustDist = 7, method = "fixed.int",
                  num.cores=2L, ignore.strand = TRUE,
                  verbose = FALSE)
DataFrame(data.frame(x2))

## The relaxed approach with method = "relaxed"
x3 = dmpClusters(GR = dmps, minNumDMPs = 6, method = "relaxed",
                  maxClustDist = 10, ignore.strand = TRUE,
                  verbose = FALSE)
DataFrame(data.frame(x3))

## ==== Setting up the experiment design to test for DMRs ===
nams <- names(dmps)
dmps_at_clusters <- getDMPatRegions(GR = dmps, regions = x3,
                                    ignore.strand = TRUE)
dmps_at_clusters <- uniqueGRanges(dmps_at_clusters, columns = 2L,
                                ignore.strand = TRUE, type = 'equal',
                                verbose = FALSE)
colnames(mcols(dmps_at_clusters)) <- nams

colData <- data.frame(condition = factor(c('CT', 'CT', 'CT',
                                        'TT', 'TT', 'TT'),
                                        levels = c('CT', 'TT')),
                    nams, row.names = 2)

## Build a RangedGlmDataSet object
ds <- glmDataSet(GR = dmps_at_clusters, colData = colData)

## ================ Testing to detect DMRs ===========
dmrs <- countTest2(DS = ds, num.cores = 4L, minCountPerIndv = 4,
                maxGrpCV = c(1, 1), Minlog2FC = 0.5,
                CountPerBp = 0.001, test = 'LRT',
                verbose = TRUE)
dmrs

## End(Not run)

genomaths/MethylIT documentation built on Feb. 3, 2024, 1:24 a.m.