dPCA: Differential Principal Component Analysis

In qwang-big/irene: R interface to interactive epigenetic analysis

Description

Run differential principal component analysis with incorporated dPCA program

Usage

dPCA(meta, bed, data, sampleId = NULL, groups = 1:2,
    datasets = NULL, transform = NULL, normlen = NULL,
    minlen = 50, lambda = 0.15, fun = function(x)
    sqrt(mean(x^2)), datasetLabels = NULL, groupLabels =
    NULL, qnormalize = TRUE, qnormalizeFirst = FALSE,
    normalize = FALSE, verbose = FALSE, interactive =
    FALSE, useSVD = FALSE, saveFile = FALSE, processedData
    = TRUE, removeLowCoverageChIPseq = FALSE,
    removeLowCoverageChIPseqProbs = 0.1, dPCsigns = NULL,
    nPaired = 0, nTransform = 0, nColMeanCent = 0,
    nColStand = 0, nMColMeanCent = 1, nMColStand = 0,
    dSNRCut = 5, nUsedPCAZ = 0, nUseRB = 0, dPeakFDRCut =
    0.5)

Arguments

meta	Meta information consists three columns (1) filenames, (2) biological groups, and (3) dataset IDs. Can be either data.frame or matrix.
bed	Input genomic regions in data.frame or matrix format, which needs to have four columns in exact this order: chromosome names, start position, end position, ID.
data	Data.frame or matrix which contains the intensity data for each input genomic regions. Columns are sorted in the order of filenames of meta information, can be produced using importBW.
sampleId	Vector of sample IDs used to be tested, equivalent to the line numbers of the meta information.
groups	Vector or List of group IDs used to be tested, specified in the group fields of the meta information. If the input is a list, for example list(1:3,4), all IDs in the first vector from the list are considered as the same group.
datasets	Vector of dataset IDs used to be tested, specified in the datasets fields of the meta information.
transform	Vector of dataset IDs which need to be transformed, power-transformations are applied according to the lambda estimated by boxcox function. If lambda is equal to or less than 0, log-transformations are applied instead.
normlen	Vector of datasets IDs which need to normalized according to the length of the genomic regions.
minlen	Numeric value of minimum length of the genomic regions to be tested, genomic regions less than this threshold are discarded.
normalize	Logical whether data will be normalized to the total library size, default is FALSE.
qnormalize	Logical whether quantile will be applied to all samples, default is TRUE.
fun	Transformation function to be applied to PCs, which produces another PCx field in the output incorporating one or several PCs. Here are some examples: function(x) sqrt(mean(x^2)) function(x) log2(mean(2^x)) function(x) x[1] function(x) abs(x[1]) function(x) x[1]-x[2] function(x) x[1]-x[2]-x[3] function(x) x[1]+x[2] function(x) sum(x) function(x) sum(abs(x)) function(x) x[1]/x[2] function(x) max(abs(x)) function(x) x[which.max(abs(x))] and by default, function(x) sqrt(mean(x^2)) is used.
verbose	Logical whether additional information and figures are shown during power-transformation and normalization.
processedData	Logical whether the processed data will be returned with other outputs.

Details

This function filters, normalizes and transforms the desired groups and datasets of the data, then forwards the processed data to a incorporated C program called dPCA (see PMID: 23569280), and outputs the dPCs and other information from the program.

Value

A named list contains three data.frames:

• gr The predefined regions followed by the computed dPCs from dPCA, as well as PCx produced by the transformation function. If processedData is TRUE, the processed data will also be append to the data.frame.

• Dobs The D matrix, which contains the observed differences between the two conditions. This is the data analyzed by dPCA.

• proj Estimated beta coefficients for each dPC.

Author(s)

Qi Wang

Examples

data(CLL)
j <- c(1,2,6,8)
res <- dPCA(meta, bed, data, groups=1:2, datasets=j, transform=j, normlen=j, processedData=TRUE, verbose=TRUE)

qwang-big/irene documentation built on May 23, 2019, 1:47 p.m.