In the counts tables made by featureCounts, the number of counts associated with each gene is always a whole number. However, in the combined normalized counts table generated by DESeq2, the counts are no longer whole numbers. How is DESeq changing the counts data, and why is this an important step in differential gene expression analysis?
the normalization method used by Deseq2 is called DESeq2’s median of ratios. This method relies in the calculation of size factors (s), whose objective is to render counts from different samples, which may have been sequenced to different depths, comparable. Herein lies the importance of normalization performed by Deseq2. Deseq2 normalized values are the result of dividing the raw counts by sample’s normalization factor (size factor, s).
You can find an extended explanation here.