How does "Normalization by a pooled sample from group" work?

[Rach]

How does “Normalization by a pooled sample from group” work?
I have 6 QC samples and each QC value changed after applying ‘Normalization by a pooled sample from group’.

Thank you in advance!

[jeff.xia]

Hi Rach, thank you for asking the question.

The method works as below:

1. Compute a reference sample based on your selected group by averaging across each feature / variable ;
2. For each sample,
   a. Compare with this reference sample to obtain a most probable dilution factor (median of the ratio);
   b. Divide each feature by this dilution factor;

Note the reference sample is the same for all, but the dilution factors are sample specific.

In Step #2, unless your QC samples contain identical values (which are unlikely), their values will still remain different. Hopefully the QC profiles will become more similar to each other (i.e. tighter cluster in PCA) compared to the profiles before normalization.