I used WGCNA to construct co-expression network based on RNA_seq data from 18 samples. I got around 60 modules. Half of the modules have less than 100 genes. I selected several modules and plot the heatmap using the expression values. I found that nearly all of the modules I checked have two expression patterns. See one example below. 

My questions: is it possible that I can adjust the parameters to separate this module into two different ones? How to do that? Or it's better I do the separation myself? Thanks. 

I'm not sure I understand your question well, but if you get two opposite expression patterns in one module, chances are it's because you have constructed an unsigned network. Rerun WGCNA using a signed or 'signed hybrid' network. I recommend 'signed hybrid' with the same soft thresholding power as your original analysis.

See http://www.peterlangfelder.com/signed-or-unsigned-which-network-type-is-preferable/ for a slightly more in-depth discussion.

HTH,

Peter