- Introduction
- Installation and running instructions for individual callers
- Build the Docker image
- Run the Docker image
- Output
Ensemble Fusion (EnFusion) merges fusion output data from Arriba, CICERO, FusionMap, FusionCatcher, JAFFA, MapSplice, and STAR-Fusion.
To learn more about this approach, visit our BioRxiv article: Discovery of Clinically Relevant Fusions in Pediatric Cancer
This list can be found here: SCRIPTS/R/GenePairCounts_2021-08-05.tsv
This is an internally generated list from the The Steve and Cindy Rasmussen Institute for Genomic Medicine at Nationwide Children's Hospital. This list contains fusion partner frequencies collected from de-identified RNA-seq data from our Comprehensive Cancer Protocol which is comprised of pediatric tumors (including central nervous system tumors, solid tumors, and hematologic malignancies). The utilization of this list allows for filtering out of recurrent and likely artifactual fusions. We will update this list biannually and will timestamp it by its release date. The below running instructions describe its use as a filtering mechanism.
Installation instructions: Installation instructions available at the Arriba GitHub: https://github.com/suhrig/arriba
Publication: Uhrig, S., et al. (2021) Accurate and efficient detection of gene fusions from RNA sequencing data. Genome Research. manuscript link
Running instructions: We utilize default settings for Arriba
Version used in publication: v1.2.0
Latest version validated: v1.2.0
Installation instructions: Installation instructions available at the CICERO GitHub: https://github.com/stjude/CICERO
Publication: Tian, L., et al. (2020) CICERO: a versatile method for detecting complex and diverse driver fusions using cancer RNA sequencing data. Genome Biology. manuscript link
Running instructions: We utilize default settings for CICERO
Version used in publication: v0.3.0
Latest version validated: v0.3.0
Installation instructions: FusionMap is a part of the Oshell toolkit and installation instructions are available at: http://www.arrayserver.com/wiki/index.php?title=Oshell#OmicScript_for_FusionMap
Publication: Ge, H., et al. (2011) FusionMap: detecting fusion genes from next-generation sequencing data at base-pair resolution. Bioinformatics. manuscript link
Running instructions: We utilize default settings for FusionMap
Version used in publication: v mono-2.10.9
Latest version validated: v mono-2.10.9
Installation instructions: Installation instructions available at the FusionCatcher GitHub: https://github.com/ndaniel/fusioncatcher
bioRxiv Preprint: Nicorici, D., et al. (2014) FusionCatcher – a tool for finding somatic fusion genes in paired-end RNA-sequencing data. bioRxiv. preprint link
Running instructions: We utilize default settings for FusionCatcher
Version used in publication: v0.99.7c
Latest version validated: v0.99.7c
Installation instructions: Installation instructions available at the JAFFA GitHub: https://github.com/Oshlack/JAFFA/wiki/Download
Publication: Wang, K., et al. (2010) JAFFA: High sensitivity transcriptome-focused fusion gene detection. Genome Medicine. manuscript link
Running instructions: We utilize default settings for JAFFA direct
Version used in publication: direct v1.09
Latest version validated: direct v1.09
Installation instructions: Installation instructions available at the MapSplice GitHub: https://github.com/LiuBioinfo/MapSplice
Publication: Davidson, N.M., et al. (2015) MapSplice: Accurate mapping of RNA-seq reads for splice junction discovery. Nucleic Acids Research. manuscript link
Running instructions: We utilize default settings for MapSplice
Version used in publication: v2.2.1
Latest version validated: v2.2.1
Installation instructions: Installation instructions available at the STAR-Fusion GitHub: https://github.com/STAR-Fusion/STAR-Fusion/wiki
Publication: Haas, B.J., et al. (2019) Accuracy assessment of fusion transcript detection via read-mapping and de novo fusion transcript assembly-based methods. Genome Biology. manuscript link
Running instructions: STAR-Fusion parameters were altered to reduce the stringency setting for the fusion fragments per million reads (FFPM) to 0.02.
--min_FFPM 0.02
Version used in publication: v1.6.0
Latest version validated: v1.6.0
We tested the instructions on Mac. They may work on Linux or Windows with or without variation.
- git
- Docker
Use git to clone the repo or navigate to the EnFusion GitHub page, download, and unzip the code.
git clone https://github.com/nch-igm/EnFusion.gitBefore building the image, if the user would like to use a known fusion list, please upload this and save it as known_fusion_list.txt in the SCRIPTS directory. Fusion partners listed in this file are not filtered out, regardless of level of support for the fusion. The known fusion list file should be a txt file where each fusion pair (gene names ordered alphabetically and genes separrated by a +) is on a single line. No header should be added. Follow example known_fusion_list.txt file (located in SCRIPTS)
Navigate to the EnFusion directory and run Docker build.
cd EnFusion
docker build . -t enfusionGet a help message from the entrypoint.
docker run enfusion -hUseful flags
--rmdeletes the container after is stops running. You can use the commanddocker container ls --allto view stopped containers that have not been deleted.-itallows for an interactive session.--entrypoint "/bin/bash"overwrites the entrypoint with the bash binary.
Please note that for the overlap script to run properly, there is an expected file structure hierarchy (refer to how test data is stored in the test_data directory for an example, however, as long as the below files are included within the mounted volume to the Docker, the overlap script will recursively search all directories under the input_location for fusion detection result files. The default search looks for the following expected file names (which are automatically generated by the individual fusion detection algorithms):
arriba = fusions.tsv
cicero = annotated.fusion.txt
fusionCatcher = final-list_candidate-fusion-genes.txt
fusionMap = FusionDetection.FusionReport.Table.txt
jaffa = jaffa_results.csv
mapSplice = fusions_well_annotated.txt
starFusion = star-fusion.fusion_predictions.abridged.tsv
We must first mount our host directory that contains the fusion detection results as a volume to our Docker container:
To do this this with the test data, save test the data directory to your local machine and then mount it as a volume:
Test data is located here: EnFusion/test_data/test
The test data contains output from 5 callers, and upon downloading this data, the directory structure should look like this:
├── test_data
├── samples
└── test
├── fusioncatcher
│ └── final-list_candidate-fusion-genes.txt
├── fusionmap
│ └── results
│ └── FusionDetection.FusionReport.Table.txt
├── jaffa
│ └── jaffa_results.csv
├── mapsplice
│ └── fusions_well_annotated.txt
└── starfusion
└── star-fusion.fusion_predictions.abridged.tsv
In this folder we also have a samples file which lists all samples included in the test_data directory. The sample listed in samples matches to the name of the other subdirectory test. More than 1 sample can be in the samples directory, as long as those same samples are included as directories (samples). In this case above, the Patient ID is test_data and the Sample ID is test.
docker run -v /~localpath/EnFusion/test_data:/SCRIPTS/test_data enfusion -o SCRIPTS/test_data/test -s test_data -p test -f 0.2
The -v flag will mount a host directory as a data volume to the docker container.
The first part (before the :) /~localpath/EnFusion/test_data needs to be the absolute path to where you have your data.
The second part (after the :)/SCRIPTS/test_data is where the data will be written to within the Docker container.
The following arguments are passed to the overlap script:
-o output_location
-s sample_ID
-p patient_ID
-f frequency cutoff to use (default is 0.10 ~ 10%, enter a decimal value here or leave blank for 10% cutoff)
Note that the -f argument utilizes the "recurrent fusion list" which has been generated from an internal RNA-seq cohort and is used to identify recurrent and likely artifactual fusions. This list can be found here SCRIPTS/R/GenePairCounts_2021-08-05.tsv and will be updated biannually and timestamped by date of collection. Please note that if you would like to prevent any filtering, you can set this argument to 0.
When running the test data you will see the following on your screen:
SCRIPTS/test_data/test
test_data
/working_dir
cp: cannot stat '/SCRIPTS/test_data/test/arriba/fusions.tsv': No such file or directory
cp: cannot stat '/SCRIPTS/test_data/test/cicero/annotated.fusion.txt': No such file or directory
[1] "/working_dir"
[1] "/SCRIPTS/R"
Assembing a list of files to merge...
[1] "starFusion" "fusionMap" "fusionCatcher" "jaffa"
[5] "mapSplice"
Because we only have 5 outputs in the test data, you will see an error messages about the missing data from arriba and cicero. This does not affect the ability of the overlap analysis to run, and is instead a note to let the user know that output from all 7 tools was not provided. For the overlap algorithm to run, output is only required from 2 callers (at minimum). The list of files to merge lets the user know which outputs were identified (of the 7 possible).
Next you will see a list of all unique fusions identified:
[1] "KIF5B+RET" "ALK+EML4" "ETV6+NTRK3" "LMNA+NTRK1"
[5] "FGFR3+TACC3" "NCOA4+RET" "NTRK1+TPM3" "PAX8+PPARG"
[9] "BRAF+SLC45A3" "BAIAP2L1+FGFR3" "ROS1+SLC34A2" "CD74+ROS1"
[13] "ERG+TMPRSS2" "EGFR+SEPTIN14" "PLEC+TSPAN4" "NEK1+SNX25"
[17] "TASOR+UBE2K" "SNX29+TXNDC11" "SAMD5+SASH1" "CBX3+CCDC32"
[21] "NAIP+OCLN" "NCOR2+UBC" "STK3+VPS13B"
Then, printed to screen are the overlap results:
[1] "# Sample: test_data\n# NumToolsAggregated: \t5\n# - starFusionCalls = \t37\n# - fusionMapCalls = \t32\n# - fusionCatcherCalls = \t355\n# - jaffaCalls = \t491\n# - mapSpliceCalls = \t26\n# filtered_overlap = \t14"
UnorderedFusion OrderedFusion KnownFusion NumTools GenePairFrequency
1 BAIAP2L1+FGFR3 FGFR3>>BAIAP2L1 yes 5 0.0103675777568332
3 CD74+ROS1 CD74>>ROS1 yes 5 0.00942507068803016
6 ETV6+NTRK3 ETV6>>NTRK3 yes 5 0.0141376060320452
7 KIF5B+RET KIF5B>>RET yes 5 0.0103675777568332
8 LMNA+NTRK1 LMNA>>NTRK1 yes 5 0.00754005655042412
9 NCOA4+RET NCOA4>>RET yes 5 0.0113100848256362
10 NTRK1+TPM3 TPM3>>NTRK1 yes 5 0.0103675777568332
11 PAX8+PPARG PAX8>>PPARG yes 5 0.0131950989632422
12 ROS1+SLC34A2 SLC34A2>>ROS1 yes 5 0.00942507068803016
2 BRAF+SLC45A3 SLC45A3>>BRAF yes 4 0.00848256361922714
5 ERG+TMPRSS2 TMPRSS2>>ERG yes 3 0.00188501413760603
4 EGFR+SEPTIN14 EGFR>>SEPTIN14 yes 2 0
The above print to screen can be quite long.
You will also see any errors or warning printed to screen.
overlap_$sample_name.tsv: full list of all intersecting fusion across callers, no filtering of any kind
filtered_overlap_knownfusionlist_3callers_$sample_name.tsv: full list of all fusions passing filter, each line for each fusion is associated with the contributing caller
collapse_filtered_overlap_knownfusionlist_3callers_$sample_name.tsv: collapsed list, where each row is associated with a single fusion (not a single caller) this output does not include all individual caller information
Singleton_KnownFusions_$sample_name.tsv: this output includes any fusions on the known fusion list that were only identified by a single caller