I have metabolomics data from a non-targeted approach normalized with internal standards. I have calculated the p-value (t-test), VIP score, and fold change.
For that, I transformed the data (each replicate) into a log 10 scale and performed the t-test (to get the p-value) and PLDSA (to get the VIP score).
I have to calculate the FOLD CHANGE (based on log 2). So, should I use the original or log-transformed data to calculate the fold change?
The difficulty in using the log-transformed data is that I have some values below one, which are converted to negative when taking the log10. Therefore, taking the ratio of treatment/non-treatment creates confusion. Though, Log (x+1) transformation could be a possible resolution to that.
Hi Jyoti,
to me the presentation of fold changes as log2 fold changes making sense because you will look for both possible direction (up or down) in an equal way. For example, if you have an up regulation which is doubled (i.e. from 1 to 2 (x2)) in log2 FC it would be 1. Whereas for a downregulation to a half (1 to 0.5 (/2) the log2 FC would be -1. You will see the benefit if you present all datqa (up and downregulated) in a volcano plot (-log 10 p-values over log2 fold change) and then try to present the same with -log 10 p-values over FC values. You will than clearly see the benefit of using Log data for fold change presentations. Dont be confused of negative log values they can be quite helpful if you want to compare up and downregulations. i.e. 1 or -1 would be a FC factor of 2 in both directions, 2 or -2 would be a FCb factor of 4, 3 and -3 an FC factor of 8, etc.
I hope this explanation helps.
Best,
Murat
- Badges
- Science topic