Hello everyone,
Normally I use TPM for within-sample analysis. Recently I got a suggestion to use Median and Quantile between-sample normalization methods. I noticed that DESeq and Limma packages offer the methods. But... what are they doing? What is the intuition behind them?
Thank you all,
Right! So more explanation:
I have a bunch of RNA-Seq experiments and I am performing some simple gene expression comparisons between two conditions (wildtype vs. mutants). Some conditions have at most two replicates. I already TPM normalized all the samples, but for comparisons, another between-sample normalization step seems like a good idea.
So far your explanations are of great help. Thank you all!
Please use
ADD COMMENT/ADD REPLY
when responding to existing posts to keep threads logically organized.If you are performing differential expression with DESeq2 or limma, don't transform the data. DESeq2 expects raw counts. For RNAseq with limma, you have to perform the
voom
transformation on the raw counts as well. Repeating: start with raw counts, not TPM, for both packages (and edgeR, for that matter).