I did quantification of RNAseq (12 samples) data using salmon. Further used tximport (either keeping countsFromAbundance="lengthScaledTPM" or “no”). Subsequently performed DESeq2 and extracted normalized data as: normalized_dds<-counts(dds, normalized=TRUE).

To check if my data is normalised, I draw box plots. However, I found in the boxplots that median values of samples are quite different. [Note: I did log2 transformation, replaced –Inf values with NA, and did Na.rm=T before drawing the boxplots in ‘R’]. my boxplot is here: https://www.dropbox.com/s/d6u9lfrdjgq6k2i/boxplots.pdf?dl=0

countsFromAbundance="lengthScaledTPM" or “no” both cases I am getting the same result that the medians are not aligned at the same level. Please advise me, what goes wrong here that data is not getting normalized?

The median ratio normalization does not provide median_i (Z_ij) = median_i (Z_ik) for sample j and sample k, and genes i, and size-factor-scaled counts matrix Z.

The data is still normalized, check MA plots to convince yourself, or scatterplots of normalized counts for two samples.