CFGL: Condition-adaptive fused graphical lasso

In Yafei611/CFGL: An adaptive procedure for inferring condition-specific gene co-expression network

Description

The function jointly construct gene co-expression network for multiple class using Condition-adaptive Fused Graphical Lasso. Pairwise screening matrics are required to adjust between-condition lasso penalty.

Usage

CFGL(Y, lambda1, lambda2, btc.screening = NULL, penalize.diag = c(TRUE,
  TRUE), weights = NULL, rho = 1, rho.increment = 1, maxiter = 500,
  tol = 1e-04, truncate = 1e-05, loglik.trace = FALSE)

Arguments

Y	A list expression data which are n*p matrices. all matrices should have a same n and p.
lambda1	The tuning parameter for the graphical lasso penalty.
lambda2	The tuning parameter for the between condition group lasso penalty.
btc.screening	A list of screening matrices (p*p) for between condition penalty. Can be obtained using the function get_scr_mat. When setting as NULL, the function will perform a standard fused graphical lasso.
penalize.diag	Binary variables that determine whether lambda1 and lambda2 are applied to the diagonal of inverse matrices.
rho	Step size parameter for ADMM algorithm. Large values decrease the step size.
rho.increment	Adjustment for rho. In each ADMM iteration, rho will be updated as rho=rho*rho.increment.
maxiter	The maximum number of ADMM interactions.
tol	The criterion for ADMM convergence.
truncate	All value in the estimated inverse convenience below this number will be set to 0.
loglik.trace	Store trace of the likelihood of estimation in each iteration.
weight	Experimental features that assigning weights to each class. Leaving it as default (NULL) is suggested.

Details

Please refer An adaptive procedure for inferring condition-specific gene co-expression network

Value

CFGL produces a list that contains estimated inverse matrices and other necessary components.

• $theta The estimation of inverse matrices

• $iters The numebr of ADMM iterations

• $loglik.trace Trace of log-likelihood

Examples

x <- expr
temp <- get_scr_mat(expr1 = x[[1]],expr2 = x[[2]])
scr.mat <- temp$scr.mat
scr.n <- sum(scr.mat)
s.sl <- temp$s

temp = CFGL(expr, lambda1 = 0.0008, lambda2 = 0.0008, btc.screening = scr.mat)
theta <- temp$theta

Yafei611/CFGL documentation built on May 25, 2019, 2:23 p.m.