Hi, I am using DRIMSeq / DEXSeq-DTU workflow describe by Love at al. 2018. According to the paper, 'scaledTPM' option is chosen for normalisation (latter on 'dtuScaledTPM' option was added to the tximport package).

1) What would be better to use: 'scaledTPM' or 'dtuScaledTPM', for DRIMSeq and DEXSeq-DTU? Is there some situation where I should prefer one over the other?

2) Since DEXSeq has it's own normalisation step, Is it beneficial to use both 'scaledTPM' (or 'dtuScaledTPM') and after that normalisation provided by DEXSeq package for DTU?

Thanks

1) I prefer dtuScaledTPM as it will be pretty close to the observed counts, while accounting for transcript length for transcripts within a gene.

2) Yes, you should use the code chunk as in the workflow:

dxd <- estimateSizeFactors(dxd)
dxd <- estimateDispersions(dxd)
dxd <- testForDEU(dxd, reducedModel=~sample + exon)