You mean a filter based on the gene-gene correlations? I think the package WGCNA might contain a function to do so.

Otherwise, you can threshold the gene-gene correlation matrix and keep the giant component only:

require(igraph) #for graph handling

require(WGCNA) #for faster "cor" function

# set cutoff for the gene-gene correlations

thr

thank you sir @Marc Santolini, here i am attaching the code .CFS is a fuction which

returns subset of feature which maximizes CFS value by taking k no of feature using Pearson's correlation coefficient. 

Dear

Look the link. May be useful.

http://machinelearningmastery.com/an-introduction-to-feature-selection/

http://www.academia.edu/7871299/A_modified_Kolmogorov-Smirnov_Correlation_Based_Filter_algorithm_for_Feature_Selection

Regards, Shafagat