I'm attempting to use DESeq to perform VST and Rlog normalization to perform survival analysis of rna-seq data. 

I realize that DESeq automatically does filtering, using genefilter, when doing DEG experiments, but that isn't true with just using the two normalization steps above, correct?

Would you suggest any type of filtering that can be implemented either before or after normalization?

Thanks for your help!

hi,

The VST and rlog do not perform filtering. Yes you might consider filtering by rowSum or row variance for clustering and heatmaps.

In the vignette, we provide code for selecting the top 30 rows by highest row sum for plotting a heatmap.

Also in the plotPCA function, we take a subset of the top genes by highest row variance, see the description of 'ntop' argument in ?plotPCA

You can get the row variances with:

rv = apply(assay(rld), 1, var)