Discrepancy between results_per_junction.tsv and results.tsv files

[scharf-f]

Dear authors,

upon our verification efforts of the aberrantSplicing module of DROP on our data I encountered discrepancies between the two mentioned output files. In the documentation you are stating about the results.tsv file: "This text file contains only significant junctions that meet the cutoffs defined in the config file, aggregated at the gene level. Any sample/gene pair is represented by only the most significant junction."
By that definition I would not expect to find calls in genes which are not present in the results_per_junction.tsv file. However, this is exactly what I encounter (actually quite regularly).

Interestingly, in two cases those are variants in positive controls we do in fact expect to see. We would miss them looking at the results_per_junction.tsv-file.

I also observed the opposite case in which I can not recover calls in genes in the results_per_junction.tsv that are listed in the results.tsv file (which I would expect to do based on the definition).

I would be very grateful for a clarification!

Also, I wanted to make sure that I get it correctly that the results listed in the html report are equivalent to the results.tsv file - is this correct?

Thank you very much,
Flo