median normalization for proteomics

Question

I am using the data from a proteomics study were the data was log2 transformed and then a median normalization was applied. The data was normalized by groups of conditions (normal, mutant), not for all the arrays at the same time, so the median for the normal group is slightly higher than for the mutant group. I am wondering if this is correct or should I normalized the data en general, I mean taking into account all the arrays and normalizing all at the same time so the 2 groups have the same median.

Regards

Answer 1

The kind of normalization depends on what you what to explore. So, there is no absolute answer to this. Different normalizations highlight different aspects of your dataset. There is a nice paper for a higher throughput quantitative study that I would recommend reading:

Nusinow, David P., John Szpyt, Mahmoud Ghandi, Christopher M. Rose, E. Robert McDonald 3rd, Marian Kalocsay, Judit Jané-Valbuena, et al. 2020. “Quantitative Proteomics of the Cancer Cell Line Encyclopedia.” Cell 180 (2): 387–402.e16.

What you want to aim for is a normalization that removes batch effects and that clusters together your quality control samples that you know are very similar to each other.