Does Fusioncatcher filter results based on supporting reads?

[paulwittgenstein]

Hi ndaniel, We are working on a research project on fusion genes. After comparing the results of several fusion-finding programs, Fusioncatcher turned out to be the most sensitive. However, we noticed that there are no results in the final-list_candidate-fusion-genes.txt output file that have less than three total supporting reads (Spanning_pairs + Spanning_unique_reads).

We ran fusioncatcher on five samples of healthy humans (our project does not focus on cancer). The same also holds true for the test output you made public here. (https://github.com/ndaniel/fusioncatcher/blob/master/test/final-list_candidate-fusion-genes.txt)

Using the arriba program (https://github.com/suhrig/arriba), we however observe fusions with only 1 or 2 supporting reads in the same samples. This is the case despite arriba being less sensitive than Fusioncatcher, at least for our purposes. We determined this by comparing which of the fusion transcripts we know for certain are present in the sample the individual programs found.

This lead us to the question: Does Fusioncatcher filter results based on supporting reads? And in general, are there any filtering steps applied that are not explicitly described in manual.md Section 9 - Methods?

Thank you for you help in advance!