mcia: multiple co-inertia analysis

Description

The main function in omicade4. Performing multiple co-inertia analysis on a list of data.frames or matrix

Usage

mcia(df.list, cia.nf = 2, cia.scan = FALSE, nsc = T, svd = TRUE)
"plot"(x, axes = 1:2,  sample.lab = TRUE, sample.legend = TRUE, sample.color = 1,  phenovec = NULL, df.color = 1, df.pch = NA, gene.nlab = 0, ...)

Arguments

df.list

A list of data.frames, matrix or ExpressionSet is going to be analyzed, the column number must be the same and mapped across all data.frame/matrix

cia.nf

An integer indicating the number of kept axes

cia.scan

A logical indicating whether the co-inertia analysis eigenvalue (scree) plot should be shown so that the number of axes, (cia.nf) can be selected interactively. Default value is FALSE.

nsc

A logical indicating whether multiple co-inertia analysis should be performed using multiple non-symmetric correspondence analyses dudi.nsc. The default =TRUE is highly recommended. If FALSE, COA dudi.coa will be performed on the first data.frame, and row weighted COA dudi.rwcoa will be performed on the rest ones using the row weights from the first one.

svd

A logical indicates which function should be used to perform singular value decomposition.

sample.lab

A logical indicating if the samples should be labelled, the default is TRUE.

sample.color

Defining colours of samples for plotting sample space, the length of this argument should be either one (uniform color) or the same with the column number of data.frame in df.list.

sample.legend

A logical indicating if the legend for sample space should be drawn.

df.color

Defining the colours for plotting variables (genes) from different data.frame. The length of this argument should be either one (all datasets use the same colour) or the same number of datasets (each dataset has a specified colour, the repetitive use of colour code is allowed.)

df.pch

Defining the pch for plotting variable (gene) space. The default is NA, the function will distinguish datasets by default. Otherwise, the length of this argument should be either one (all datasets use the same pch) or the same number of datasets (each dataset has a specified pch).

phenovec

A factor for plotting sample space, phenovec could be used to distinguish individuals in the data.frames.

An object of class mcia

axes

A vector of integer in length 2 to indicate the axes are going to be plotted. The default are first two axes.

gene.nlab

An integer indicating how many top weighted genes on each axis should be labelled

...

Other arguments

Value

call: the function called
mcoa: The results returned by mcoa
coa: The results returned by separate analysis (applying dudi.nsc or dudi.coa on each data.frame separately)

Details

The column number of data.frame in the df.list must be the same, and the same column from different data.frame should be matchable. For example, Microarray profiling for the same set of cell lines, patients and etc.

mcia calls dudi.nsc, ktab and mcoa in ade4 packages.

Plotting and visualizing mcia results

Two functions could be used to visualize the result of mcia: The first is plot.mcia, which results in four plots. Top left represents the sample space. Individuals from the same column of different data.frames are linked by edges. Different platforms are distinguished by the shape of points. Top right shows the variable space, datasets are marked by different colours. Bottom left represents the eigenvalue scree plot. The pseudo-eigenvalue space of all data.frames are visualized in the bottom right panel. The second function is plotVar.mcia, which could be used to plot the variable space for different datasets as well as finding and visualizing the variables (genes) across datasets.

Other methods

selectVar.mcia: selecting variables (genes) according to the their coordinates.

Examples

Run this code


data(NCI60_4arrays)
mcoin <- mcia(NCI60_4arrays)
plot(mcoin, sample.lab=FALSE, df.col=4:7)

colcode <- sapply(strsplit(colnames(NCI60_4arrays$agilent), split="\\."), 
                  function(x) x[1])
plot(mcoin, sample.lab=FALSE, sample.color=as.factor(colcode))