| Title: | Modelling Experimental Data from MeDIP Enrichment |
|---|---|
| Description: | MEDME allows the prediction of absolute and relative methylation levels based on measures obtained by MeDIP-microarray experiments |
| Authors: | Mattia Pelizzola and Annette Molinaro |
| Maintainer: | Mattia Pelizzola <[email protected]> |
| License: | GPL (>= 2) |
| Version: | 1.65.0 |
| Built: | 2024-07-03 05:50:35 UTC |
| Source: | https://github.com/bioc/MEDME |
The count of CpGs is determined in each window of size wsize, with or withouth weighting, for each probe according to its position, chromosome and genome realease
CGcount(data, wsize = 1000, wFunction = "linear")CGcount(data, wsize = 1000, wFunction = "linear")
data |
An object of class MEDMEset |
wsize |
number; the size of the smoothing window, in bp |
wFunction |
string; the type of weighting function, to choose among linear, exp, log or none |
Only human and mouse are currently supported. The respective genomic sequence metadata library needs to be downloaded from the Bioconductor website, installed and loaded (around 800Mb). Please note that only the last genome release should be used. LiftOver UCSC tool could be used for batch conversion of old genomic position to the last genome release.
An object of class MEDMEset is returned where the count of CpGs for each probe has been saved on the CGcount slot.
data(testMEDMEset) ## just an example with the first 1000 probes testMEDMEset = smooth(data = testMEDMEset[1:1000,]) library(BSgenome.Hsapiens.UCSC.hg18) testMEDMEset = CGcount(data = testMEDMEset)data(testMEDMEset) ## just an example with the first 1000 probes testMEDMEset = smooth(data = testMEDMEset[1:1000,]) library(BSgenome.Hsapiens.UCSC.hg18) testMEDMEset = CGcount(data = testMEDMEset)
Probe-level MeDIP weighted enrichment is compared to the expected DNA methytlation level. The former is determined applying MeDIP protocol to a fully methylated DNA. The latter is determined as the count of CpGs for each probe. This is assumed to be the methylation level of each probe in a fully methylated sample.
MEDME(data, sample, CGcountThr = 1, figName = NULL)MEDME(data, sample, CGcountThr = 1, figName = NULL)
data |
An object of class MEDMEset |
sample |
Integer; the number of the sample to be used to fit the model, based on the order of samples in the smoothed slot |
CGcountThr |
number; the threshold to avoid modelling probes with really low methylation level, i.e. CpG count |
figName |
string; the name of the file reporting the model fitting |
The model should be applied on calibration data containing MeDIP enrichment of fully methylated DNA, most likely artificially generated (see references). Nevertheless, in case chromosome or genome-wide human tiling arrays are used a regular sample could be used too. In fact, human genomic DNA is known to be hyper-methylated but in the promoter regions. Of course the performance of the method is expected to be somehow affected by this approximation.
The logistic model as returned from the multdrc function from the drc R library
http://genome.cshlp.org/cgi/content/abstract/gr.080721.108v1
data(testMEDMEset) ## just an example with the first 1000 probes testMEDMEset = smooth(data = testMEDMEset[1:1000, ]) library(BSgenome.Hsapiens.UCSC.hg18) testMEDMEset = CGcount(data = testMEDMEset) MEDMEmodel = MEDME(data = testMEDMEset, sample = 1, CGcountThr = 1, figName = NULL)data(testMEDMEset) ## just an example with the first 1000 probes testMEDMEset = smooth(data = testMEDMEset[1:1000, ]) library(BSgenome.Hsapiens.UCSC.hg18) testMEDMEset = CGcount(data = testMEDMEset) MEDMEmodel = MEDME(data = testMEDMEset, sample = 1, CGcountThr = 1, figName = NULL)
This allows the probe-level determination of MeDIP smoothed data, as well as absolute and relative methylation levels (AMS and RMS respectively)
MEDME.predict(data, MEDMEfit, MEDMEextremes = c(1,32), wsize = 1000, wFunction='linear')MEDME.predict(data, MEDMEfit, MEDMEextremes = c(1,32), wsize = 1000, wFunction='linear')
data |
An object of class MEDMEset |
MEDMEfit |
the model obtained from the MEDME.model function |
MEDMEextremes |
vector; the background and saturation values as determined by the fitting of the model on the calibration data |
wsize |
number; the size of the smoothing window, in bp |
wFunction |
string; the type of weighting function, to choose among linear, exp, log or none |
An object of class MEDMEset. The resulting smoothed data, the absolute and relative methylation score (AMS and RMS) are saved in the smoothed, AMS and RMS slots, respectively.
http://genome.cshlp.org/cgi/content/abstract/gr.080721.108v1
data(testMEDMEset) ## just an example with the first 1000 probes testMEDMEset = smooth(data = testMEDMEset[1:1000, ]) library(BSgenome.Hsapiens.UCSC.hg18) testMEDMEset = CGcount(data = testMEDMEset) MEDMEmodel = MEDME(data = testMEDMEset, sample = 1, CGcountThr = 1, figName = NULL) testMEDMEset = MEDME.predict(data = testMEDMEset, MEDMEfit = MEDMEmodel, MEDMEextremes = c(1,32), wsize = 1000, wFunction='linear')data(testMEDMEset) ## just an example with the first 1000 probes testMEDMEset = smooth(data = testMEDMEset[1:1000, ]) library(BSgenome.Hsapiens.UCSC.hg18) testMEDMEset = CGcount(data = testMEDMEset) MEDMEmodel = MEDME(data = testMEDMEset, sample = 1, CGcountThr = 1, figName = NULL) testMEDMEset = MEDME.predict(data = testMEDMEset, MEDMEfit = MEDMEmodel, MEDMEextremes = c(1,32), wsize = 1000, wFunction='linear')
allows to read sgr or gff files before submitting the data to MEDME analysis
MEDME.readFiles(path = getwd(), files = NULL, format, organism)MEDME.readFiles(path = getwd(), files = NULL, format, organism)
path |
string; the path where the files are stored; the current working directory is the default |
files |
vector; optional vector of file names |
format |
string; either sgr or gff to indicate the respective file formats |
organism |
string; either hsa or mmu for homo sapiens and mus musculus respectively |
In case of GFF files (recommendend), tab-delimited files with header are expected with following fields: chromosome, probe ids, start and stop chromosomal positions, and score are expected in columns 1, 3, 4, 5 and 6 repectively. Multiple files are also expected to be in the same order of rows.
In case of sgr files (GFF is the preferred format), tab-delimited files with no header and chr, chr positions and score are expected in columns 1, 2 and 3 repectively. Multiple files are also expected to be in the same order of rows.
An object of class MEDMEset. The column headers in the logR slot are determined from the file names. in case of SGR files the are not probe names and progressive numbers are used in place of them. In case of GFF files the probe names are determined from the 3rd column.
allows to write sgr or gff files after MEDME analysis
MEDME.writeFiles(data, output, path = getwd(), format, featureLength = NULL)MEDME.writeFiles(data, output, path = getwd(), format, featureLength = NULL)
data |
An object of class MEDMEset |
output |
string; the name of the data slot to be written on the disk, either logR, smoothed, AMS or RMS |
path |
string; the path where the files are stored; the current working directory is the default |
format |
string; either sgr or gff to indicate the respective file formats |
featureLength |
integer; in case of GFF file format the length of the features has to be provided to determine start and end positions |
One GFF or SGR file is provided for each sample of the data MEDMEset object.
In case of GFF files, tab-delimited files with header are provided with following fields for each probe: chromosome, empty field, probe ids, start and stop chromosomal positions, and score and empty fields.
In case of sgr files, tab-delimited files with no header and chr, chr positions and score are provided.
This class is used in MEDME library to store MeDIP derived DNA-methylation estimates and to save further elaboration of these, in association with chromosomal and positional probe information
Objects can be created by calls of the form new("MEDMEset", ...).
This object could initially host the MeDIP normalized logRatio data, as returned by the MEDME.readFiles function. Afterwards,
the same obejct is returned by most of the MEDME library function. Each time, a new slot is filled with additional data, as smoothed
logR or Absolute/Relative Methylation Scores (AMS and RMS respectively). At the end of the analysis, usually after a call to the
MEDME.predict function, the MEDME.writeFiles function can be used to generate SGR or GFF files from this object.
chr:Object of class "character" : the probe-level chromosome asignments
pos:Object of class "numeric" : the probe-level genomic position
logR:Object of class "matrix" : the probe-level un-trasformed normalized MeDIP logRatios for each sample
smoothed:Object of class "matrix" : the probe-level smoothed MeDIP logRatios for each sample
AMS:Object of class "matrix" : the probe-level Absolute Methylation Score for each sample
RMS:Object of class "matrix" : the probe-level Relative Methylation Score for each sample
CGcounts:Object of class "numeric" : the probe-level count of CpGs
organism:Object of class "character" : the organism that the probe genomic positions are referring to, either hsa or mmu for homo sapiens or mus musculus respectively
signature(x = "MEDMEset"): subsets the object based on its probes and/or samples
signature(object = "MEDMEset"): extracts the Absolute Methylation Score from the AMS slot
signature(object = "MEDMEset"): extracts the probe CpG count from the CGcounts slot
signature(object = "MEDMEset"): extracts the probe chromosomal assignment
signature(object = "MEDMEset"): extracts the organism
signature(.Object = "MEDMEset"): automatically generates smoothed, AMS and RMS matrix when only the logR slot is filled
signature(object = "MEDMEset"): extracts the matrix of MeDIP un-transformed logRatios
signature(object = "MEDMEset"): extracts the probe genomic position
signature(object = "MEDMEset"): extracts the Relative Methylation Score from the RMS slot
signature(object = "MEDMEset"): prints a summary of the object content
signature(object = "MEDMEset"): extracts the Absolute Methylation Score from the AMS slot
Mattia Pelizzola
http://genome.cshlp.org/cgi/content/abstract/gr.080721.108v1
MEDME.readFiles, MEDME.writeFiles
showClass("MEDMEset")showClass("MEDMEset")
MeDIP data from tiling arrays are smoothed by determining for each probe i the weighted average of the probes within a window of size wsize centered at i
smooth(data, wsize=1000, wFunction='linear')smooth(data, wsize=1000, wFunction='linear')
data |
An object of class MEDMEset |
wsize |
number; the size of the smoothing window, in bp |
wFunction |
string; the type of weighting function, to choose among linear, exp, log or none |
The un-smoothed data are read from the slot logR of the data MEDMEset and the resulting smoothed data are saved on the smoothed slot.
An object of class MEDMEset. In particular, the smoothed data are saved on the smoothed slot.
data(testMEDMEset) # just an example with the first 1000 probes testMEDMEset = smooth(data = testMEDMEset[1:1000,])data(testMEDMEset) # just an example with the first 1000 probes testMEDMEset = smooth(data = testMEDMEset[1:1000,])
This dataset contains a subset of the data reported in references. It contains normalized un-smoothed probe-level MeDIP enrichment for almost 50000 probes. This is a random subset of a custom Nimblegen chromosome X tiling array. It is a two channels array with an resolution of 100bp and oligos of 60nt. The M value is reported only. The fullyMet column of the logR slot contains data from a calibration experiments where MeDIP has been applied to a fully methylated sample. The last two columns NBMEL and YUSAC2 contain DNA-methylation experimental data for two cell strains: NBMEL are newborn normal melanocytes cells and YUSAC2 a melanoma strain. Data was processed with within and between array normalization. The full dataset contains almost 380K probes. See references for details. Chromosome, genomic position and logR of probes can be accessed with the methods chr, pos and logR respectively.
Please note that the original genomic coordinates were mapped to the hg17 human genome. These have been converted to hg18 using the LiftOver USCS tool available online for batch conversion.
data(testMEDMEset)data(testMEDMEset)
MEDMEset
http://genome.cshlp.org/cgi/content/abstract/gr.080721.108v1