Hi, I'm doing an internship where one the goal is the discovery of rare and/or novel transcripts. After reading the docs especially the StringTie's one I remarked that the pipeline does provide the possibility of de novo assembly of transcripts through

--stringtie_ignore_gtf

however it does not follow StringTie's author recommended workflow where the reads and the reference genome go through StringTie a first time for assembly then the assembled transcripts are merged with the reference to form a new GTF which will be then use for transcripts abundance's and coverage's. In Pertea et al., 2016 they explain the necessity to merge because reads may only partially cover an unknown transcripts and produces partial versions if not merged. I'm not the only one who have made this observation because some user already commented this but without opening a PR (https://github.com/nf-core/rnaseq/issues/858). Is there a particular reason for the current workflow rather than the recommended one ? If not I would be interested to make the necessary modifications.

This was added before my time on the pipeline, but suspect it's just historical, and to be honest it hasn't been widely used functionality in the pipeline. Feel free to PR and we can go from there.