I recently acquired a metatranscriptomics dataset. There are ~1 million ORFs and after strict filtering could be reduced to ~100 thousand ORFs. The dataset has treatment (N=49) and control (N=34) groups for a total of 83 samples.

I have access to compute resources which I can use for WGCNA.  

I only need WGCNA for the following:

1. picking a soft threshold; 
2. TOM matrix calculation
3. dendrogram cutting

I've seen the blockwise modules method but this seems strange in having multiple incomparable dendrograms.  Is the blockwise modules method the recommended approach for a dataset of this size?  What are some other approaches for running WGCNA on a dataset this large?