regarding coexpression analysis (by pearson method) of genes shortlisted from expression profile of a dataset obtained from GDAC Firehouse database

Hello guys,
Should i consider genes having “zero” normalized count given in expression data of a dataset while doing coexpression analysis with respect to a particular gene (whose expression value is positive). I have taken data from Broad gdac firehouse database..its mRnaseq (illuminahiseq_rnaseqv2-RSEM_genes_normalized (MD5)) data, obtained from database.