relationship between DE-genes and DE-trascripts in deseq2

Question

[eyonesi]

Hello everybody, I ran a DE –RNAseq project by star-stringtie2- deseq2 pipline. The prepDE.py was used for generation of transcript matrix and gene matrix. Deseq2 were performed at both gene and transcript levels. At the gene level, 838 DE-genes were identified with -10< logFC< +10, it seems that this results are ok. But at the transcript level, just 112 DE-transcripts were identified with high logFC (-30<logFC < +30). Since, the transcripts are part of the genes; I don’t understand why the number of transcripts is less than the number of genes and why their logFC is higher? What is relationship between gene expression and transcript expression at deseq2? Is genes expression calculated based on the mean expression of their transcripts?

Regards

[swbarnes2]

DESeq isn't designed to be run with transcripts, but with genes, so I don't think your transcript results are valid.

Answer 1

[rpolicastro]

In general you don't want to directly pass transcript counts to DESeq2 due to the uncertainty of assigning fragments to transcripts causing an increased false discovery rate. Instead you should use a workflow that accounts for this, such as salmon + fishpond or kallisto + sleuth.