scoring.boot.parallel function

Calculating differential score with parallel bootstrap scoring

This function calculates standard errors for edge-wise partial correlation differences obtained from DINGO model. Bootstrapping is done in parallel using parSapply from the "parallel" library.

scoring.boot.parallel(stddat,z,Omega,A,B,boot.B=100,verbose=T,cores=1)

Arguments

• stddat: standardized nxp data with colnames as genename
• z: a length n vector representing a binary covariate
• Omega: a p x p precision matrix for std dat which implies the global network
• A: p x p matrix of the MLE for the baseline covariance matrix which is obtained from A value of the Greg.em function.
• B: p x 2 matrix of the MLE for the regression coefficient which is obtained from B value of the Greg.em function
• boot.B: a scalar indicating the number of bootstraps
• verbose: if TRUE, lists the bootstrap replications
• cores: the number of cores to run in parallel for bootstrapping, set to 1 as a default.

Returns

• genepair: a p(p-1)/2 x 2 matrix indicating all pairs of genes

• levels.z: a length 2 vector indicating levels of the binary covariate z, the first element is for group 1 and the second element is for group 2

• R1: a length p(p-1)/2 vector indicating partial correlations for group 1 and the order is corresponding to the order of genepair

• R2: a length p(p-1)/2 vector indicating partial correlations for group 2 and the order is corresponding to the order of genepair

• boot.diff: a p(p-1)/2 x boot.B matrix indicating bootstrapped difference, Fisher's Z transformed R1 - R2. The rows are corresponding to the order of gene pair and the columns are corresponding to the bootstrap samples

• diff.score: a p(p-1)/2 vector of differential score corresponding to genepair

• p.val: a p(p-1)/2 vector of corrected p-values corresponding to genepair

Author(s)

Min Jin HA mjha@mdanderson.org, Caleb CLASS caclass@mdanderson.org