Hello, I am using tximport to generate txi list from upstream kallisto output.

Then I use the txi list to construct a SummarizedExperiment object.

Here is the code for constructing the se object:

se <- SummarizedExperiment(assays = list(counts = txi$counts, 
abundance = txi$abundance, 
length = txi$length), 
rowData = rownames(txi$counts), 
colData = metadata)

So should I use the counts file for the downstream DESeq2 analysis?

> assays(se)
List of length 3
names(3): counts abundance length

This is covered extensively in the tximport vignette.