impulseDE2 input

Question

[Omics data mining]

Hello everyone

I am using impulseDE2 for the longitudinal analysis. I found this link https://issueexplorer.com/issue/YosefLab/ImpulseDE2/28 which indicates not to use normalized count and decided to use raw count data for the analysis. I also found computeSizeFactors (https://rdrr.io/bioc/ImpulseDE2/man/computeSizeFactors.html) function within the impulseDE2 which means by default it will calculate sizefactor for normalization.

But I have seen many post where they mention to use normalized count from DESeq2. I am bit confused about the "input" file for impulseDE2.

I would appreciate all the suggestions.

Answer 1

[kelen]

If I am not mistaken impulseDE2 does indeed require raw counts as it does it's own DE, including using DESeq2 for some steps:

"ImpulseDE2 is based on a negative binomial noise model with dispersion trend smoothing by DESeq2 and uses the impulse model to constrain the mean expression trajectory of each gene" - from here https://bioc.ism.ac.jp/packages/3.9/bioc/manuals/ImpulseDE2/man/ImpulseDE2.pdf

The results of significantly DE genes were roughly the same as they were with DESeq2 LRT, when I tried it a while ago, but I didn't fully understand everything it was doing so I opted not to use it.