knorm {knorm}R Documentation

Microarray Data From Multiple Biologically Interrelated Experiments

Description

Produces Knorm correlations between genes (or probes) from microarray data obtained across multiple biologically interrelated experiments. The Knorm correlation adjusts for experiment dependencies (correlations) and reduces to the Pearson coefficient when experiment dependencies are absent. The Knorm estimation approach can be generally applicable to obtain between-row correlations from data matrices with two-way dependencies.

Usage

knorm(data, bsamples, thres_diff, thres_ev1, thres_ev2, burn_in, no_subgenes, no_fullgenes, repli)

Arguments

data matrix containing (normalized) gene expression data. Rows correspond to arrays and columns correspond to genes (or probes). Data from replicates of experiments are placed in consecutive rows.
bsamples number of bootstrap samples for estimation.
thres_diff threshold of difference between log-likelihood values.
thres_ev1 threshold for eigen values of experiment covariance matrix. Eigen values smaller than thres_ev1 are considered negligible.
thres_ev2 threshold for eigen values of gene covariance matrix. Eigen values smaller than thres_ev2 are considered negligible.
burn_in minimum number of iterations for estimation. Default is 2.
no_subgenes number of genes (or probes) to be used in the row-subsampling technique for estimating the experiment covariance matrix. This number should not be more than the number of experiments in the data.
no_fullgenes number of genes (or probes) in data.
repli vector of number of replicates for each experiment. For example, c(2,3) denotes two replicates for experiment 1 and three replicates for experiment 2.

Details

This estimation procedure consists of a gene (or row) sub-sampling and a covariance shrinkage technique that iteratively estimates the gene and experiment covariance matrices. The covariance shrinkage method using the diagonal matrix with unequal covariances as the target matrix was used (Schafer and Strimmer, 2005). For more details on the estimation procedure, model assumptions and conditions, please refer to Teng et al. (2007).

Value

A list containing:

a_cor_est Experiment correlation matrix estimate.
g_cor_est Knorm correlations (between genes).
m_est Mean matrix estimate.

Author(s)

Siew Leng Teng

References

Teng, S.L., Huang, H., and Zhou, X. Jasmine. (2008), "A statistical framework to infer functional gene relationships from biologically interrelated microarray experiments"

Examples

#Importing simulated Multiple Microarray data.
#For more information on data set imported, look at help file for mmcd
#for futher information.
data(mmcad)

#Creating vector fo the number of replicates for each experiment.  There
#will be three replications for each experiment in the mmcd data.
repli=rep(3,30)

results <- knorm(mmcad, 25, 0.01, 1e-10, 1e-10, 2, length(repli),ncol(mmcad),repli)
a_cor_est <- results$a_cor_est
g_cor_est <- results$g_cor_est

[Package knorm version 1.0 Index]