RNA-seq Co-expression network construction from the output of DESeq2

Question

[pixie@bioinfo]

Hello, I have generated a read count_matrix using DESeq2. I would like to construct co-expression network using WGCNA. For microarray data, they suggest to use log-transformed (base2) data.

For RNA-seq data, what options do I have in the DESeq package ? What normalization methods should I follow ? Ant other input is greatly appreciated.

Answer 1

[Jake Warner]

Hi!
Point 4 in the WGCNA FAQ addresses this:

We then recommend a variance-stabilizing transformation. For example, package DESeq2 implements the function varianceStabilizingTransformation which we have found useful, but one could also start with normalized counts (or RPKM/FPKM data) and log-transform them using log2(x+1). For highly expressed features, the differences between full variance stabilization and a simple log transformation are small.

Good luck!

[pixie@bioinfo]

I should have read that part :) Thanks for pointing out!

[Jake Warner]

No problem, try it a few ways. I found little difference between log2(x+1), rlog and varianceStabilizingTransformation when it came to the most highly connected genes.

[pixie@bioinfo]

Thank you, I will try them out. One small query, do you do any filtering for genes with very low expression values before the transformations ? What are the typical cutoffs you use ?

[Jake Warner]

Yes I filter lowly expressed features. The developers of WGCNA recommend as well:

Probesets or genes may be filtered by mean expression or variance (or their robust analogs such as median and median absolute deviation, MAD) since low-expressed or non-varying genes usually represent noise.

DEseq2 has some automatic filtering step so you should be able to recover that as your filtering criterion. See "Automatic Independent Filtering":
https://genomebiology.biomedcentral.com/articles/10.1186/s13059-014-0550-8