Fits dependency models to chromosomal arm, chromosome or the whole genome.
1 2 3 4 5 6 7 8 9 10 11 12 13 14
Data sets 1. It is recommended to place gene/mirna expression data in X and copy number data in Y. Each is a list with the following items: data: in a matrix form. Genes are in rows and samples in columns. e.g. gene copy number; info: Data frame which contains following information aboout genes in data matrix; chr: Number indicating the chrosome for the gene: (1 to 24). Characters 'X' or 'Y' can be used also.; arm: Character indicating the chromosomal arm for the gene ('p' or 'q'); loc: Location of the gene in base pairs; pint.data: can be used to create data sets in this format.
The second data item.
Determine the window size. This specifies the number of nearest genes to be included in the chromosomal window of the model, and therefore the scale of the investigated chromosomal region. If not specified, using the default ratio of 1/3 between features and samples or
Specify the chromosome for model fitting. If missing, whole genome is screened.
Specify chromosomal arm for model fitting. If missing, both arms are modeled.
Dependency screening can utilize any of the functions from the package dmt (CRAN). Particular options include 'pSimCCA' probabilistic similarity constrained canonical correlation analysis Lahti et al. 2009. This is the default method.; 'pCCA' probabilistic canonical correlation analysis Bach & Jordan 2005; 'pPCA' probabilistic principal component analysis Tipping & Bishop 1999; 'pFA': probabilistic factor analysis Rubin & Thayer 1982; 'TPriorpSimCCA' probabilistic similarity constrained canonical correlation analysis with possibility to tune T prior (Lahti et al. 2009)
To be used with segmented data, or nonmatched probes in general. Using nonmatched features (probes) between the data sets. Development feature, to be documented later.
Regularization by nonnegativity constraints on the projections. Development feature, to be documented later.
List of parameters for the dependency model.
sigmas: Variance parameter for the matrix normal prior
distribution of the transformation matrix T. This describes the
deviation of T from H
H: Mean parameter for the matrix normal prior distribution
prior of transformation matrix T
zDimension: Dimensionality of the latent variable
mySeed: Random seed.
covLimit: Convergence limit. Default depends on the selected
method: 1e-3 for pSimCCA with full marginal covariances and 1e-6
for pSimCCA in other cases.
max.dist: Maximum allowed distance between probes. Used in
automated matching of the probes between the two data sets based on
chromosomal location information.
outputType: Specifies the output type of the function. possible values are
Logical. Determines the useage of the method for segmented data
screen.cgh.mrna assumes that data is already
paired. This can be done with
pint.match. It takes sliding
gene windows with
fixed.window and fits dependency models
to each window with
fit.dependency.model function. If the
window exceeds start or end location (last probe) in the chromosome in
fixed.window function, the last window containing the
given probe and not exceeding the chromosomal boundaries is used. In
practice, this means that dependency score for the last n/2 probes in
each end of the chromosome (arm) will be calculated with an identical
window, which gives identical scores for these end position probes. This
is necessary since the window size has to be fixed to allow direct
comparability of the dependency scores across chromosomal windows.
screen.cgh.mir calculates dependencies
around a chromosomal window in each sample in
X; only one sample
X will be used. Data sets do not have to be of the same size
X can be considerably smaller. This is used with e.g. miRNA
data. If method name is specified, this overrides the corresponding model
parameters, corresponding to the modeling assumptions of the specified
model. Otherwise method for dependency models is determined by
parameters. Dependency scores are plotted with dependency score plotting.
The type of the return value is defined by the the function argument
outputType. With the argument
outputType = "models", the return value depends on the other arguments; returns a ChromosomeModels which contains all the models for dependencies in chromosome or a GenomeModels which contains all the models for dependencies in genome. With the argument
outputType = "data.frame", the function returns a data frame with eachs row representing a dependency model for one gene. The columns are: geneName, dependencyScore, chr, arm, loc
Leo Lahti firstname.lastname@example.org
Dependency Detection with Similarity Constraints, Lahti et al., 2009 Proc. MLSP'09 IEEE International Workshop on Machine Learning for Signal Processing, See http://www.cis.hut.fi/lmlahti/publications/mlsp09_preprint.pdf
A Probabilistic Interpretation of Canonical Correlation Analysis, Bach Francis R. and Jordan Michael I. 2005 Technical Report 688. Department of Statistics, University of California, Berkley. http://www.di.ens.fr/~fbach/probacca.pdf
Probabilistic Principal Component Analysis, Tipping Michael E. and Bishop Christopher M. 1999. Journal of the Royal Statistical Society, Series B, 61, Part 3, pp. 611–622. http://research.microsoft.com/en-us/um/people/cmbishop/downloads/Bishop-PPCA-JRSS.pdf EM Algorithms for ML Factoral Analysis, Rubin D. and Thayer D. 1982. Psychometrika, vol. 47, no. 1.
To fit a dependency model:
ChromosomeModels holds dependency models for chromosome,
GenomeModels holds dependency models for genome. For
dependency score plotting
Add the following code to your website.
For more information on customizing the embed code, read Embedding Snippets.