Big difference in DEGs number from different tools

,
1

Hello to everyone,

i am trying to analyze my RNAseq Data. So i follow the tutorial:

Bérénice Batut, Mallory Freeberg, Mo Heydarian, Anika Erxleben, Pavankumar Videm, Clemens Blank, Maria Doyle, Nicola Soranzo, Peter van Heusden, 2020 Reference-based RNA-Seq data analysis (Galaxy Training Materials) . /training-material/topics/transcriptomics/tutorials/ref-based/tutorial.html

Just briefly: so i have my paired-end FASTQ files
I use STAR aligner with built-in MM10 Full genome, without gene model (nothing selected)
To calculate counts i used fetureCounts with buit-in MM10 or use .gtf file with Release M25 (GRCm38.p6)… Already here i have a bit different count numbers (built-in vs M25).

To identify DEGs i tried several tools. With DESeq2 i have more than 400 DEGs (FDR<0.05)
If i use EdgeR or limma-trend, then i have only few DEGs with FDR or adjusted p-values <0.05

So my question is why the number of DEGs are so extreamly different in different tools?
And another question if i need to specify gtf file during maping with STAR?

Thank you for your answers and help.

2

You need to make sure that you are using annotations that correspond to the genome you mapped against. So if you mapped against mm10, you must use annotations for mm10. There will always be some variation among different tools. The variation you are seeing is likely not what you would expect. Are you sure that you used equivalent parameter setting across the three tools?

3

nekrut, thank you (spasibo bol’shoe ;-))

So if i use builtin MM10 genome in RNA STAR, then I can not use this annotation file:
ftp://ftp.ebi.ac.uk/pub/databases/gencode/Gencode_mouse/release_M25/gencode.vM25.annotation.gtf.gz

Where can I get then the appropriate annotation file?

About Parameters in the tools I am not really sure, since in EdgeR and limma i can filter low counts, so i filter on cpm.

In DESeq2 there is no filtering function, so most probably huge difference come from here. But at the same time there are a lot of places where i read, that it is not necessary to filter low counts for DESeq2…

So i am just confused and trying to undesratnd it :wink: