Question about normalization method before differential expressed gene (DEG) analysis.

[YCHuang0610]

Dear all, I want to identify differential expressed genes (DEGs) between ROIs (for example, PFC versus the rest of brain) using limma R package, after parcellation through the abagen.get_expression_data function. I am wondering which normalization method in {sample,gene}_norm will be recommend before DEG analysis, since the static test is sensitive to the data distribution.

What's more, when I used the 'srs' normalizaiton method by default, the distribution of different ROI in each donors seemed to be weird:

expr = abagen.get_expression_data(imagemask, data_dir=datadir, donors="9861",
                                  donor_probes='independent', probe_selection='corr_intensity')
sns.boxplot(data=expr.T)

And when the norm method was set to None:

expr = abagen.get_expression_data(imagemask, data_dir=datadir, donors="9861", gene_norm=None, sample_norm=None,
                                  donor_probes='independent', probe_selection='corr_intensity')

The parcellation mask was shown bellow:

Thank you for your help, Yc Huang.