Can VSN normalized values be treated the same as log-transformed values?

Analysis
1

I have proteomics data across two experimental conditions that I normalized using the normalize_met() function, which I understand does VSN normalization. I’ve done further downstream analysis including PLS-DA and using VIP scores and loading weights to get my list of “significant” proteins.

If I want to get log2 fold changes of the original abundances of my “significant” proteins, is it sufficient to treat the VSN normalized values similar to log-transformed values and take the difference in values as the ratio?

Or do I need to just go back to the raw abundance matrix and apply log-transform.

Thank you.

2

Hi @padak

I won’t be able to provide specific comments as normalisation is outside mixOmics. My general comment is that the proteins of interest are defined according to their VSN normalised values and so this is how you should plot / interpret them afterwards (you can’t really reverse-engineer, and you need to acknowledge that the data have been transformed first before going through data analysis). Of course you can decide to change the normalisation first before analysis, this is up to you.

Kim-Anh