attractorScanningGL: Find all genomically-localized attractors in the dataset

In weiyi-bitw/cafr: Attractor Metagenes Finding Algorithm

Description

Exhaustively search for all converged genomically-localized attractors in the dataset.

Usage

attractorScanningGL(data, genome, alpha=(2:12)/2, windowSize = 50, maxIter = 100, epsilon=1E-14, bin=6, so=3, score.position=5, num.output=10, negateMI=TRUE, verbose=TRUE, saveAllScore=FALSE)

Arguments

data	An expression matrix with genes in the rows, samples in the columns.
genome	A matrix with genome information. The rownames of the matrix must contains gene symbols in their genomical order. See data(genome) for an example.
alpha	A vector of exponents to try for convergence.
windowSize	Number of genes to be filtered in, centered at the seed gene.
maxIter	Max number of iterations.
epsilon	Threshold of convergence.
bin	Number of bins used when estimate mutual information (default=6).
so	Spline order used when estimate mutual information (default=3).
score.position	Which rank in the attractor will be used as the strength. (default=5)
num.output	Number of genes to include in the attractor.
negateMI	When TRUE, negate the mutual information if the two vectors have negative momentum.
verbose	When TRUE, it shows the progress.
saveAllScore	When TRUE, the function output every converged attractor using each seed. It could take a lot of memory.

Details

The genomically-localized attractor usually points to an amplicon or delecon that contains several genes being amplified or deleted simultaneously. Such amplification and deletion within the region reflects on the co-expression of genes in the neighborhood. attractorScanningGL exhaustively uses every gene as a seed in the dataset to find all converged genomically locallized attractors in the region of the seed gene. It uses the same algorithm as findAttractor, but restricts the gene space to the neighborhood of the seed gene defined by the windowSize. The significance (or strength) of an attractor is defined by the MI of the score.position-th gene. Default is the fifth-highest MI in the attractor. The function tries all the possible exponents defined in alpha and finds the best one that gives highest score.position-th MI. It then compares all the converged attractors using the seeds in the neighborhood and finds the one with the highest strength.

Value

When saveAllScore is TRUE, it returns a list with the following fields:

attractome	A matrix of all converged attractors. Each column is the genes of the attractor ranked by their MI with the metagene. Each column name is the seed used to generate the attractors.
bestAlphas	The best exponents that lead to the highest strength given the seed.
score	The strength of each attractor.
scoremat	A matrix of the MIs corresponding to each gene in the attractome field.
summary	The summarized output after removing the overlapping attractors. The output only contains the attractor with highest strength among all overlapping ones. Each row is the genes of an attractor ranked according to their MIs with the metagene. Each row name is the seed that leads to the attractor. The first column gives the cytoband information. The last column gives the strength of the attractor.

If saveAllScore is FALSE, the function only output the summary field above in matrix form.

Author(s)

Wei-Yi Cheng

References

Wei-Yi Cheng, Tai-Hsien Ou Yang and Dimitris Anastassiou, Biomolecular events in cancer revealed by attractor metagenes, PLoS Computational Biology, Vol. 9, Issue 2, February 2013.

See Also

findAttractor, findGLAttractor, attractorScanning

Examples

## Not run: 
# Load the toy dataset extracted from TCGA OV data
data(ov)

# Load the genome information
data(genome)

# find every genomically localized attractor in the toy dataset
out <- attractorScanningGL(ov, genome)

# display the summarized output
print(out)

## End(Not run)

weiyi-bitw/cafr documentation built on May 4, 2019, 4:18 a.m.