Filter RNA-seq matrix

Question

[zhang.jianhai]

Dear community,

I am using RNA-seq data for gene network analysis. In data pre-processing, I filter out genes with low expresion counts and genes with low coefficient of variance (CV). I found the CV on counts and logs-transformed values are quite different, which means the former is significantly larger than the later. Thus, which is better regarding CV filtering, on counts or log2-values?

Regards.

[ATpoint]

Check what the manual of the tool you use recommends.