COSINE (version 2.0)

GA_search: Use genetic algorithm to search for the globally optimal subnetwork

Description

This function performs the stochastic search using genetic algorithm to find the globally optimal subnetwork which gives rise to the highest score defined by a scoring function, which measures the extent of the differential expression of the subnetwork across several datasets.

Usage

GA_search(lambda, diff_expr, diff_coex, num_iter = 1000, 
muCh = 0.05, zToR = 10)

Arguments

lambda

A vector containing the five quantiles of the weight parameter lambda

diff_expr

A vector storing the F-statistics measuring the differential expression of each gene, which length equals the number of genes N

diff_coex

An N by N matrix with entry (i,j) corresponding to the ECF-statistics of gene pair (i,j), which measures the differential correlation between genes i and j

num_iter

The number of iterations to be performed by the genetic algorithm

muCh

the mutation chance used by genetic algorithm

zToR

zero to one ratio

Value

• A list containing the following components:
• Subnet_size
  A vector containing the size of the subnetwork identified using each lambda
• Best_Scores
  A vector containing the best scores of the subnetworks
• Subnet
  A list containing the extracted subnetworks (a list of genes) for each of the five lambda values
• GA_obj
  A list of the returned objects of the genetic algorithm function

References

http://cran.r-project.org/web/packages/genalg/index.html

Examples

# Load the scaled F-statistics and ECF-statistics 
# for the simulated datasets

data(set1_scaled_diff)

# Get the quantiles of lambda

klist<-c(25,30)
set1_quantile<-get_quantiles(diff_expr=set1_scaled_diff[[1]], 
diff_coex=set1_scaled_diff[[2]],klist,pop_size=10)
lambda<-set1_quantile[[2]]

#Perform genetic algorithm to search-just show the first iteration here

set1_GA<-GA_search(lambda,diff_expr=set1_scaled_diff[[1]],
diff_coex=set1_scaled_diff[[2]], num_iter=1, muCh=0.05, zToR=10)