I have transcript abundance data (obtained with kallisto), which I converted into non-normalized gene-level counts with tximport. In my resulting count matrix, each gene has an Ensembl ID. Before running DESeq2, I converted the gene Ensembl IDs into gene symbols. However, it so happens that a few different Ensembl IDs map to more than one gene symbol. Thus, in my count matrix, 16 genes are duplicated. Is it ok to add the counts of the duplicated genes together for each sample?

Thank you.

I wouldn’t do this operation of collapsing to symbols before statistical analysis. Afterward you can think of various ways to integrate the evidence.