Calculating FC on probabilistic quotient normalization (PQN) data

Question

[bridget.tripp]

I am working with publicly available metabolomics data that has been probabilistic quotient normalized (PQN). I am interested in calculating the fold changes (FC) between case/control. Some of the normalized values are negative so I am unsure if I am capturing the true differences between samples when I apply more traditional FC calculations. Is there an approach you have used to calculate fold changes on these types of normalized data?