Hi @bellez34r3
Stringtie is a discovery tool, and it doesn’t call (or annotate) newly predicted coding regions in discovered transcripts.
See the process at this specific step in the tutorial you referenced. https://training.galaxyproject.org/training-material/topics/transcriptomics/tutorials/differential-isoform-expression/tutorial.html#hands-on-transcriptome-assembly-with-stringtie
The other tools are important and the options used differ between runs through Stringtie. Each sample is run through twice – once for discovery per sample, then those are merged to remove redundancy, then all samples are run through again using the merged result as a new “reference only” set. The gffread steps are important for first gathering the inputs needed to call CDS regions, then run again to actually get those regions captured in the final annotation.
If possible, maybe use the workflow included with that tutorial as a template? Or, at least some parts of it? Or you could run the tutorial data through to create a sort of “reference history” then review the tools and parameters applied, compare the different files produced, and learn where your process is different?
Let us know if you need more help or if you sort this out. It should definitely work. 