Hi,
my question is regarding the most common diff. expr. tools used (edger, deseq2, limma). As they are created before UMIs existed (where the overall set of data includes quite low copy numbers on all entries which seem like rare low expressed genes) did such data can be used with their statistical methods and normalization and get proper DEG results? I'm not a statistician but know that changing the scale of the data, dispersion, etc may cause bias in the results?
1 answer
UMI counts will work perfectly fine for these tools and should actually better match the statistical assumptions underlying them than regular read counts (due to the additional noise from PCR duplicates here). The issue is more when you get to drop-outs (missed signal), which can occur in things like scRNA-seq. In practice that still seems to be handled well enough by these tools.
Log in to answer this question.
-
RNA-seq normalization for time-dependent analysis
written by luca.sclisizzo •
-
Can I use removeBatchEffect() before performing DEG analysis in bulk RNA-seq with limma-voom?
written by Sumeet •
-
Why employ normalization methods, and how can they be utilized in DEG analysis?
written by wyt1995 •
-
the dispersion estimation of edgeR and DESeq2
written by tommy •
-
Can FPKM data sets be of any use or are they trash?
written by schulpen_91 •
-
limma - voom with TPM data
written by AlexStar •
-
heatmaps for limma-voom DE genes
written by AlexStar •
-
Normalization methods for high-throughput RNASeq (CQN or TMM)
written by shangyuan5000 •
-
RNA-Seq replicates with big differences in medians of VST values
written by panconchoclo •
-
Chip-seq analysis with input and spike-in
written by Damian Kao
