Description:
COBRA decomposes a (partial) gene co-expression matrix as a
linear combination of covariate-specific components.
It can be applied for batch correction, differential co-expression
analysis controlling for variables, and to understand the impact of
variables of interest to the observed co-expression.
Usage
cobra(X, expressionData, standardize = T)
Arguments
- X
: design matrix of size (n, q), n = number of samples, q = number of covariates
- expressionData
: gene expression as a matrix of size (g, n), g = number of genes
- standardize
: boolean flag to standardize the gene expression as a pre-processing step
Outputs:
Value
psi : impact of each covariate on the eigenvalues as a matrix of size (q, n)
Q : eigenvectors corresponding to non-zero eigenvalues as a matrix of size (g, n)
D : non-zero eigenvalues as a list of length n
G : (standardized) gene expression as a matrix of size (g, n)
Examples
g <- 100 # number of genes
n <- 10 # number of samples
q <- 2 # number of covariates
X <- X <- cbind(rep(1, n), rbinom(n, 1, 0.5))
expressionData=matrix(rnorm(g*n, 1, 1), ncol = n, nrow = g)
# Run COBRA algorithm
cobra_output <- cobra(X, expressionData)