glmnet and RNA-seq data

Question

[Anastaziel]

Hi,

I am trying to use glmnet to perform Cox analysis. I have 150 samples and for 30k genes I have gene counts.

These are the two questions I have.

1. How would I go about normalizing the gene counts to fit into the assumptions of glmnet?

2. How should I go about selecting a smaller subset of genes?

For point 1, I am thinking I can use vst (from DESeq2) or maybe voom (limma) with calcNormFactors (edgeR) but I am stuck wondering about the standardize = TRUE option from glmnet.

For point 2, I was using rowVars from DESeq2 after I had normalized my data with either vst or voom with calcNormFactors.

And so currently, I am stuck deciding what to do for points 1 and 2. Any help would be appreciated.

[Mark]

Hi, can you please checkout these two bioconductor forum posts to see if they answer your question? Both answers are from experts in the field and both are authors of edgeR/DEseq2: https://support.bioconductor.org/p/93732/ https://support.bioconductor.org/p/93160/

I think combined they will help you figure out the right way forward.