I am trying to figure out some traits using WGCNA methodology.

This analysis is based on microarray meta-analysis. Briefly explaining, I did analyze some features of differentially expressed genes from integrating independent data set (microarray data set containing profiles of patients and healthy subjects) using random effect inverse-variance methodology based on effect size.

In WGCNA procedure, I took specific value per gene by

1) calculate PCC from gene expression profiles per each data set,  
2) converting PCC into Fisher's Z-statistic from each data set, 
3) integrating different data sets using random effect inverse-variance methodology, and then 
4) converting back to correlation coefficient using the reverse Fisher's Z-statistic.

To identify co-expressed network, I considered this reversed correlation coefficient as adjacency value and follow typical R WGCNA commands.

However, I don't have any idea about the way to calculate the principle components from gene expression profiles.

Actually, all procedures to combine independent different data sets were built on gene-centric view. As you may know well, one of biggest challenges in meta-analysis is inherent and naturally-deriven variability and variance between each data set. To calculate eigengenes from raw gene expression profiles, what should I do? I've thought about the unified methodology such as z-value transformation per each data set and treat the z-value as raw gene expression profiles. However, I am not sure whether this method is relevant to follow it.