This is a test version of Biostars. For the public version, visit https://www.biostars.org.
RNA-seq normalization and differential expression

Sorry friends,

I got totally confused, as I understood in a common RNA-seq analysis, for example with tophat, after producing accepted_hits.bam, reads are counted by featurecount or another tools to output a read count matrix. Hereafter if the normalization (for example log transforming) is the second step or producing the RPKM or VST by some tools such as DESe2 is the same differential expression analysis and normalization together? I mean is normalization an independent step before differential expression analysis?

Thank you

rna-seq
• 5,232 views • link
updated by Ram Ram • written by zizigolu zizigolu

4 answers

I'd say normalization is part of differential expression analysis. Depending on the biological context of your data and how you performed the sequencing, you may need to do additional normalization or quality control. Arguably, quality control could be considered part of it too or as a preprocessing step.

• 0 views • link
written by jnf3769 • jnf3769

You pointed to quality control as pre-processing step. In my work, the initial quality control checking by drawing box plot or MA plot showed that the data are not normalized. In this situation, could you please let me know if I have to do normalization before doing differential expression analysis using say edgeR tool or it should do just on the results of differentially expressed genes to compare samples?

• 0 views • link
written by seta seta

Hi!

Most tools for differential expression analysis handle normalization by themself. In the case of DESeq2 you can see from the documentation that only raw read counts must be used :

The count values must be raw counts of sequencing reads. This is important forDESeq2's statistical model to hold, as only the actual counts allow assessing the measurement precision correctly. Hence, please do not supply other quantities, such as (rounded) normalized counts, or counts of covered base pairs this will only lead to nonsensical results.

PS : be wary that DESeq2 doesn't output RPKM values. RPKM values shouldn't be used in differential analysis anyway.

• 0 views • link
updated by Ram Ram • written by Carlo Yague Carlo Yague

thank you so much

• 0 views • link
written by zizigolu zizigolu

differential expression analysis can include the entire pipeline. Its general term. Normalisation is performed before trying to find differentially expressed genes.

• 0 views • link
updated by Ram Ram • written by GouthamAtla GouthamAtla

Thank you, for example for my understanding, is log transformation before using DESeq2?

• 0 views • link
updated by Ram Ram • written by zizigolu zizigolu
If you mean normalization by log transformation, DESeq2 won't work with normalized values.
• 0 views • link
written by cpad0112 cpad0112

DESeq and EdgeR do not perform differential expression directly on normalized values. They calculate a library size factor and use that factor in their downstream differential expression tests.

• 0 views • link
updated by Ram Ram • written by Damian Kao Damian Kao

Log in to answer this question.

More posts like this