Fusion caller OUTPUT

Each sample is called with 5 separate callers:

– chimerascan-0.4.5
– defuse-0.6.2
– fusioncatcher_v0.99.3e
– MapSplice-v2.1.8
– STAR-Fusion_v0.5.4

The output of each is stored in a separate caller specific folder: `fusion/CALLER-NAME`, e.g. `fusion/defuse`

In the top level `fusion` folder: there is a _merge_ file for each sample that merges the calls from all callers into one file for downstream analysis. The files are called: <ProjName>_merged_fusions_<SAMPLE_NAME>.txt_ranked.txt (example: Proj_12008_merged_fusions_s_18-41607.txt_ranked.txt). In addition the colums that describe the fusions from the samples there are columns that RANK the genes and break points based on how many callers call the same gene/event.

ColNum,ColName
20,TAG_genes
21,FusionMetaCaller_rank_genes
22,TAG_breakpoints
23,FusionMetaCaller_rank_breakpoints

The genes columns look for fusion between the same fusion partner genes but with potentially different breakpoints. The breakpoints columns are more specific and look for repeated calls of the

To call fusions each sample is processed with 5 fusion callers: chimerascan (v0.4.5), defuse (v0.6.2), fusioncatcher (v0.99.3e), MapSplice (v2.1.8), STAR-Fusion (v0.5.4). The output from each caller is then merged to create a union list of all events using a custom merge script that is available here: https://github.com/soccin/BIC-RNAseq/tree/master/tools/MergeFusion. Column 1: CALLER_ID indicates which caller found the event.

The merged file can be used to as is which gives a list of so called maximum sensitivity; meaning it tries to find all events at the price of having potentially many FALSE POSITIVES. If one filters for gene/events that are seen in multiple callers then when can get a list of successively more specific calls at the price of missing events; ie a higher FALSE NEGATIVE rate.

Fusion caller OUTPUT

Fusion caller pipeline