Hi,

We have a question about the proper handing of ambiguous annotation reads when using DESeq2. In our bacterial RNAseq data there is a fraction of reads (0.5-2%) that map inconclusively to two adjacent genes (gene A and gene B). We count such reads twice, one for gene A and one for gene B. Can we use DESeq2 for such counting? We prefer not to exclude these reads because they may have biological meaning.

Thank you very much, Raya

You can use a method such as Salmon to perform quantification followed by tximport and DESeqDataSetFromTximport. This is our official pipeline for taking in the proportioned reads, and has been thoroughly benchmarked.

Additionally, we have a method Swish in the fishpond Bioconductor package which focuses on when there is substantial uncertainty in read assignment, as occurs in transcript level analysis or in gene level for scRNA-Seq.