Inconsistent microarray expression levels after normalizing with log2

Question

I know it is necessary to normalize the GEO microassay data before analysing. Howerver, after using log2 to normalize the data and checking with boxplot, I found the data is still not align in the same level:

I notice the expression in some samples is lower overall, can I add a value manually to it?

Update: As @ATPoint had pointed out, log2 is not a normalization method but transformation. So it is normal to see the median number of sample data not at the same level. Actually, I need to normalize it via limma or other R package.

Answer 1

As I already told you here (When analysing microarray data, is it need to do normalization and standardization both?) log2 is a transformation, not a normalization so unless these data were already normalized you cannot expect equal medians.

What the method section tells is

"The scanned images were analyzed with Feature Extraction Software 9.5 (Agilent) using default parameters (protocol : GE1-v5_95_Feb07 and Grid: 014850_D_F_20060807) to obtain background subtracted and spatially detrended Processed Signal intensities. Features flagged in Feature Extraction as Feature Non-uniform outliers were excluded."

So that is probably what these values are. Hence, you might want to apply something like quantile normalization (for example via the limma package) to the log2-transformed data before applying limma for differential testing or any other downstream normalization. Read the limma manual. Note that I did not invest much time to read my self into that experiment. Treat advise with care and collaborate with local experienced people if you need hands-on guidance.