Hello,

I have a question regarding DEXSeq. I ran both my mapped (STAR) paired-end reads and unpaired/singleton reads (result of Trimmomatic) through the DEXSeq python scripts. As a result, I have counts for both paired and singleton data. I'm hesitant to sum the read counts, as I did with HTseq-count prior to DESeq2 since both paired-end and unpaired reads add a fragment count. Does DEXSeq count exon overlap of both the forward and reverse reads? For example, if read F1 overlaps exon 1 and read R1 overlaps exon 3, both exon 1 and exon 3 will receive a count? And does this matter then if I only have a forward singleton read that overlaps exon 1 and is missing the reverse read that would've mapped to exon 3? Perhaps it's better to solely use paired counts in the analysis, or to control for read type using read:exon in the design? Any advice would be appreciated.

Thank you, Melissa

Hi Melissa,

Some in-line answers to your questions:

Does DEXSeq count exon overlap of both the forward and reverse reads? For example, if read F1 overlaps exon 1 and read R1 overlaps exon 3, both exon 1 and exon 3 will receive a count?

The scripts count sequenced fragments (so one pair will be counted once). However, multi-overlaps are allowed if the fragments maps to exons of the same gene. In your example, both exon 1 and exon3 will receive a count.

And does this matter then if I only have a forward singleton read that overlaps exon 1 and is missing the reverse read that would've mapped to exon 3?

The default of the python scripts script will be to ignore singletons.

Perhaps it's better to solely use paired counts in the analysis, or to control for read type using read:exon in the design?

Additional variables are assigned per sample. Not sure I understand this question.

Best regards, Alejandro