An accurate tool to detect gene fusion from long transcriptomics reads.
Genion comes with a stand-alone binary and helper Snakemake to assist mapping and preparing reference files.
You can install genion through conda, docker or from source.
Installation with bioconda
conda install -c bioconda genionCheck out mamba. A faster conda reimplementation.
git clone https://github.com/vpc-ccg/genion
cd genion/docker
docker build . -t genion:latestAfter genion is built with Docker, you can run it with the following command
docker run --user=$UID -v /path/to/inputs:/input -v /path/to/outputdir:/output genion [args]| Dependencies | Version |
|---|---|
| c++ | gcc >= 9 or clang >= 8 |
| zlib | >= 1.2.11 |
git clone https://github.com/vpc-ccg/genion
cd genion
makeGenion requires following input files to run:
- Mapping file of transcriptomics long reads (paf): Genion does not do mapping. It accepts mappings in paf format. You can use any splice-aware long read to whole genome mapper (and convert sam to paf using paftools if mapper doesn't output paf).
- Long reads file(fast{a,q}): These are used for filter low complexity sequence filtering
- Gene annotation file (GTF)
- Sequence similarity file: This is used to filter candidates from genes with similar sequences. This file is produced by all to all mapping cDNA reference file with itself. It can be created using genion snakemake or command line given in the Required References section. This file is a tab separated 2 column file containing transcript pairs. This file can be produced using ENSEMBL cDNA reference and following command line:
minimap2 [cdna.fa] [cdna.fa] -X -t [threads] -2 -c -o [cdna.selfalign.paf]
cat [cdna.selfalign.paf] | cut -f1,6 | sed 's/_/\t/g' | awk 'BEGIN{OFS=\"\\t\";}{print substr($1,1,15),substr($2,1,15),substr($3,1,15),substr($4,1,15);}' | awk '$1!=$3' | sort | uniq > [cdna.selfalign.tsv]- Duplication annotation: Genomic segmental duplication annotation. This used to filter out candidates that come from copies of the same segmental duplication. For hg38, it can be downloaded from ftp://hgdownload.soe.ucsc.edu/goldenPath/hg38/database/genomicSuperDups.txt.gz.
./genion
-i /path/to/input/fastq
--gtf /path/to/annotation/gtf
--gpaf /path/to/genomic/mapping/paf
-s /path/to/gene/homology/tsv
-d /path/to/genomicSuperDups.txt
-o /path/to/output/tsv - [output]: Contains called gene fusions and readthrough. It is a tab separated sheet with the following columns.
gene1.id::gene2.id gene1.name::gene2.name ffigf-score FiN-score supporting-reads normal-counts pass-fail-code ranges- ffigf-score of fusion A::B : Number of supporting A::B fusion reads divided by number of fusion reads mapping to gene A or gene B but not both.
- FiN-score: Number of supporting A::B fusion reads divided by sum of number of normal A reads and B normal reads (normal being reads not supporting any gene fusions)
- normal-counts: ';' separated list of non-chimeric read counts of the member genes.
- pass-fail-code: PASS:GF for gene fusions, PASS:RT for readthroughs, FAIL::reason::... for filtered candidates.
- ranges: semicolon separated set of (median) genomic intervals showing which part of the member genes are expressed (sorted in the same order of gene ids)
- [output].fail: Contains called filtered fusion candidates in the same column format.
- [output].log: Contains genomic positions of the supporting reads formatted as:
read-id gene1-id gene1-range gene2-id gene2-range
You can download a small simulated example from: https://figshare.com/articles/dataset/Small_gene_fusion_simulated_long_read_dataset/17253821
example.fastq #Simulated reads fastq
example.paf #Mapping in paf format
genomicSuperDups.txt #Genomic segmental duplication annotation
Homo_sapiens.GRCh38.97.gtf #Gene annotation
cdna.self.tsv #Homology information
tar xzvf small_example.tar.gzcd small_example
genion -i example.fastq -d genomicSuperDups.txt --gtf Homo_sapiens.GRCh38.97.gtf -g example.paf -s cdna.self.tsv -t 1 -o output.tsv Upon the successful run of genion on this example dataset, output.tsv should look like this.
Additionally, we provide a snakemake file to help running genion.
- Maps Long reads
- Downloads the duplication annotation
- Prepares the Sequence similarity file
- Runs Genion
| Dependencies | Version |
|---|---|
| c++ | gcc >= 9 or clang >= 8 |
| zlib | >= 1.2.11 |
| Python | >= 3.7 |
| snakemake | >= 5.3.0 |
| deSALT | >= 1.5.5 |
| minimap2 | >= 2.17 |
| paftools |
Snakemake dependencies can be installed using conda/mamba
conda create --file genion.env --name genion-env
conda activate genion-envIn order to run Genion, you need to create a project configuration file namely config.yaml.
This configuration consists of a number mandatory settings and some optional advance settings.
Below is the list of the all the settings that you can set in your project.
| config-paramater-name | Type | Description |
|---|---|---|
| path | Mandatory | Full path to project directory. |
| reference-dna | Mandatory | Full path to the DNA reference |
| reference-cdna | Mandatory | Full path to the cDNA reference |
| annotation-gtf | Mandatory | Full path to the GTF annotation |
| rawdata-base | Mandatory | Location of the input fastq files relative to path. |
| or rawdata | Mandatory | Full path to the location of the input fastq files |
| input | Mandatory | A list of input files per sample. See the following example |
| ext | Optional | (Important, if this is wrong snakemake won't run correctly) extension of the fastq files used in input (fastq,fastq.gz,fq,fq.gz) default:fastq |
| genion-binary | Optional | Path to genion binary, should be set if genion is not in $PATH |
| desalt-index | Optional | If not provided, reference will be indexed on the run |
| analysis-base | Optional | Location of intermediate files relative to path. default: {path}/analysis |
| or analysis | Optional | Full path to the location of intermediate files. default: {path}/analysis |
| results-base | Optional | Location of final results relative to the path. default: {path}/results |
| or results | Optional | Full path to the location of final results. default: {path}/results |
| wg-aligner | Optional | Mapper to use (deSALT, minimap2) default: deSALT |
Each input requires a fastq file and type. Type is used to configure parameters by the mapper. Following are the available types of input:
| type | Technology |
|---|---|
| ccs | PacBio SMRT CCS reads: error rate 1% |
| clr | PacBio SMRT CLR reads: error rate 15% |
| ont1d | Oxford Nanopore 1D reads: error rate > 20% |
| ont2d | Oxford Nanopore 2D reads: error rate > 12% |
The following a an example of config-yaml for Nanopore and Pacbio runs for a sample
path:
/path/to/project/directory
annotation-gtf/:
/path/to/annotation/gtf
reference-dna:
/path/to/reference/dna
reference-cdna:
/path/to/reference/cdna
desalt-index:
/path/to/index/dir/
ext:
fastq.gz
wg-aligner:
deSALT
input:
"A":
type:
clr
fastq:
- A_clr.fastq.gz
"B":
type:
ont1d
fastq:
- B_ont.fastq.gz[path]/
├── rawdata
│ ├── A_clr.fastq.gz
| └── B_ont.fastq.gz
├── analysis (intermediate files)
│ ├── A
| └── B
└── results
├── A.fusions.tsv
├── A.readthrough.tsv
├── B.fusions.tsv
└── B.readthrough.tsv
For the input/output file structure description, snakemake configuration comes with two options each for rawdata, analysis and results.
You can use -base suffix (like rawdata-base). This way snakemake will know that given path is relative to the project path.
Or you can directly use rawdata to enter absolute path. This may be helpful if input files are not in the project directory.
After preparing a config file following the Project Configuration section, you can run snakemake with the following command.
snakemake -j [number-of-threads] --config-file [path-to-config-file]