StringTie reports 0 coverage for all reference transcripts

Question

[PoGibas]

I have used StringTie on mice RNA-seq data. StringTie was run with parameters given in Nature paper. Reference gtf annotation was downloaded from the Gencode Genes.
Most of the transcripts in the reference annotation have very low expression values (> 95% transcripts have 0 fpkm; for one sample all reference transcripts have 0 fpkm). However, half of newly merged transcripts have non-zero expression values. My questions are:

1. Is it normal for all the reference transcripts to have expression values equal to zero?
2. After running StringTie with default parameters I got many (maybe too many) newly merged transcripts (140K vs 25K in the reference annotation). This looks kinda strange and I would prefer to use transcripts from the reference, but as I mentioned above - they have 0 expression values.

Answer 1

[k.mooo]

0 expression values could indicate an incongruous reference genome between alignment and annotation. The same version, as well as the same source (for example, Ensembl or UCSC), must be used when aligning reads and annotating reads. This is mentioned at the bottom of this blog with respect to reference genome builds. https://github.com/griffithlab/rnaseq_tutorial/wiki/Annotation