arriba

Gene fusion

RNA-Seq

Structural variants

Chimeric alignments

Description

Arriba is a command-line tool for the detection of gene fusions from RNA-Seq data. It was developed for the use in a clinical research setting. Therefore, short runtimes and high sensitivity were important design criteria.

Arriba is based on the STAR RNA-Seq aligner and post-processes the alignments (output from STAR) to:

detect split reads and discordant mates, which are indicative of structural rearrangements,
find reads supporting gene fusions (i.e., reads spanning the breakpoints of gene fusions),
perform various filtering steps to remove false positives, and
output the final predictions in a standardized format.

In contrast to many other fusion detection tools, Arriba does not require to reduce the STAR parameter --alignIntronMax (maximum intron size). Reducing this parameter impairs detection of long introns and may affect expression quantification. Arriba reliably filters translocation-based false positives even when large maximum intron sizes are used.

Important: Arriba requires BAM files that were aligned with STAR using specific chimeric alignment parameters, particularly --chimOutType WithinBAM HardClip. See the official workflow documentation for the complete set of recommended STAR parameters.

Type

bash_script

License

MIT

Keywords

Gene fusion

RNA-Seq

Structural variants

Chimeric alignments

Links

Repository

Documentation

Contributors

Launch

Run this component

Run the following command to execute this component with Nextflow:

cat > params.yaml <<'EOM'  
fusions: "$id.$key.fusions.tsv"  
fusions_discarded: "$id.$key.fusions_discarded.tsv"  
id: "run"  
publish_dir: "output/"  
EOM

nextflow run https://packages.viash-hub.com/vsh/biobox.git \  
  -revision v0.4.1 \  
  -main-script target/nextflow/arriba/main.nf \  
  -params-file params.yaml

Inputs

Name	Type & Properties	Description
--bam -x	file required	File in SAM/BAM/CRAM format with main alignments as generated by STAR (Aligned.out.sam). Important: The BAM file must be generated by STAR with chimeric alignment parameters, specifically `--chimOutType WithinBAM HardClip`. Arriba extracts candidate reads from this file, including both normal and chimeric alignments.
--genome -a	file required	FastA file with genome sequence (assembly). The file may be gzip-compressed. An index with the file extension .fai must exist only if CRAM files are processed.
--gene_annotation -g	file required	GTF file with gene annotation. The file may be gzip-compressed.
--known_fusions -k	file	File containing known/recurrent fusions. Some cancer entities are often characterized by fusions between the same pair of genes. In order to boost sensitivity, a list of known fusions can be supplied using this parameter. The list must contain two columns with the names of the fused genes, separated by tabs.
--blacklist -b	file	File containing blacklisted events (recurrent artifacts and transcripts observed in healthy tissue).
--structural_variants -d	file	Tab-separated file with coordinates of structural variants found using whole-genome sequencing data. These coordinates serve to increase sensitivity towards weakly expressed fusions and to eliminate fusions with low evidence.
--tags -t	file	Tab-separated file containing fusions to annotate with tags in the 'tags' column. The first two columns specify the genes; the third column specifies the tag. The file may be gzip-compressed.
--protein_domains -p	file	File in GFF3 format containing coordinates of the protein domains of genes. The protein domains retained in a fusion are listed in the column 'retained_protein_domains'. The file may be gzip-compressed.

Name

Type & Properties

--bam

-x

file

required

--genome

-a

file

required

--gene_annotation

-g

file

required

--known_fusions

-k

file

--blacklist

-b

file

--structural_variants

-d

file

--tags

-t

file

--protein_domains

-p

file

Outputs

Name	Type & Properties	Description
--fusions -o	file required output	Output file with fusions that have passed all filters.
--fusions_discarded -O	file output	Output file with fusions that were discarded due to filtering.

Name

Type & Properties

--fusions

-o

file

required

output

--fusions_discarded

-O

file

output

Arguments

Name	Type & Properties	Description
--max_genomic_breakpoint_distance -D	long	When a file with genomic breakpoints obtained via whole-genome sequencing is supplied via the --structural_variants parameter, this parameter determines how far a genomic breakpoint may be away from a transcriptomic breakpoint to consider it as a related event. For events inside genes, the distance is added to the end of the gene; for intergenic events, the distance threshold is applied as is. Default: 100000.
--strandedness -s	string	Whether a strand-specific protocol was used for library preparation, and if so, the type of strandedness (auto/yes/no/reverse). When unstranded data is processed, the strand can sometimes be inferred from splice-patterns. But in unclear situations, stranded data helps resolve ambiguities. Default: auto
--interesting_contigs -i	string multiple	List of interesting contigs. Fusions between genes on other contigs are ignored. Contigs can be specified with or without the prefix "chr". Asterisks () are treated as wild-cards. Default: 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 X Y AC_ NC_*
--viral_contigs -v	string multiple	List of viral contigs. Asterisks () are treated as wild-cards. Default: AC_ NC_*
--disable_filters -f	string multiple	List of filters to disable. By default all filters are enabled.
--max_e_value -E	double	Arriba estimates the number of fusions with a given number of supporting reads which one would expect to see by random chance. If the expected number of fusions (e-value) is higher than this threshold, the fusion is discarded by the 'relative_support' filter. Note: Increasing this threshold can dramatically increase the number of false positives and may increase the runtime of resource-intensive steps. Fractional values are possible. Default: 0.300000
--min_supporting_reads -S	integer	The 'min_support' filter discards all fusions with fewer than this many supporting reads (split reads and discordant mates combined). Default: 2
--max_mismappers -m	double	When more than this fraction of supporting reads turns out to be mismappers, the 'mismappers' filter discards the fusion. Default: 0.800000
--max_homolog_identity -L	double	Genes with more than the given fraction of sequence identity are considered homologs and removed by the 'homologs' filter. Default: 0.300000
--homopolymer_length -H	integer	The 'homopolymer' filter removes breakpoints adjacent to homopolymers of the given length or more. Default: 6
--read_through_distance -R	integer	The 'read_through' filter removes read-through fusions where the breakpoints are less than the given distance away from each other. Default: 10000
--min_anchor_length -A	integer	Alignment artifacts are often characterized by split reads coming from only one gene and no discordant mates. Moreover, the split reads only align to a short stretch in one of the genes. The 'short_anchor' filter removes these fusions. This parameter sets the threshold in bp for what the filter considers short. Default: 23
--many_spliced_events -M	integer	The 'many_spliced' filter recovers fusions between genes that have at least this many spliced breakpoints. Default: 4
--max_kmer_content -K	double	The 'low_entropy' filter removes reads with repetitive 3-mers. If the 3-mers make up more than the given fraction of the sequence, then the read is discarded. Default: 0.600000
--max_mismatch_pvalue -V	double	The 'mismatches' filter uses a binomial model to calculate a p-value for observing a given number of mismatches in a read. If the number of mismatches is too high, the read is discarded. Default: 0.010000
--fragment_length -F	integer	When paired-end data is given, the fragment length is estimated automatically and this parameter has no effect. But when single-end data is given, the mean fragment length should be specified to effectively filter fusions that arise from hairpin structures. Default: 200
--max_reads -U	integer	Subsample fusions with more than the given number of supporting reads. This improves performance without compromising sensitivity, as long as the threshold is high. Counting of supporting reads beyond the threshold is inaccurate, obviously. Default: 300
--quantile -Q	double	Highly expressed genes are prone to produce artifacts during library preparation. Genes with an expression above the given quantile are eligible for filtering by the 'in_vitro' filter. Default: 0.998000
--exonic_fraction -e	double	The breakpoints of false-positive predictions of intragenic events are often both in exons. True predictions are more likely to have at least one breakpoint in an intron, because introns are larger. If the fraction of exonic sequence between two breakpoints is smaller than the given fraction, the 'intragenic_exonic' filter discards the event. Default: 0.330000
--top_n -T	integer	Only report viral integration sites of the top N most highly expressed viral contigs. Default: 5
--covered_fraction -C	double	Ignore virally associated events if the virus is not fully expressed, i.e., less than the given fraction of the viral contig is transcribed. Default: 0.050000
--max_itd_length -l	integer	Maximum length of internal tandem duplications. Note: Increasing this value beyond the default can impair performance and lead to many false positives. Default: 100
--min_itd_allele_fraction -z	double	Required fraction of supporting reads to report an internal tandem duplication. Default: 0.070000
--min_itd_supporting_reads -Z	integer	Required absolute number of supporting reads to report an internal tandem duplication. Default: 10
--skip_duplicate_marking -u	boolean_true	Instead of performing duplicate marking itself, Arriba relies on duplicate marking by a preceding program using the BAM_FDUP flag. This makes sense when unique molecular identifiers (UMI) are used.
--extra_information -X	boolean_true	To reduce the runtime and file size, by default, the columns 'fusion_transcript', 'peptide_sequence', and 'read_identifiers' are left empty in the file containing discarded fusion candidates (see parameter -O). When this flag is set, this extra information is reported in the discarded fusions file.
--fill_gaps -I	boolean_true	If assembly of the fusion transcript sequence from the supporting reads is incomplete (denoted as '...'), fill the gaps using the assembly sequence wherever possible.

Name

Type & Properties

--max_genomic_breakpoint_distance

-D

long

--strandedness

-s

string

--interesting_contigs

-i

string

multiple

--viral_contigs

-v

string

multiple

--disable_filters

-f

string

multiple

--max_e_value

-E

double

--min_supporting_reads

-S

integer

--max_mismappers

-m

double

--max_homolog_identity

-L

double

--homopolymer_length

-H

integer

--read_through_distance

-R

integer

--min_anchor_length

-A

integer

--many_spliced_events

-M

integer

--max_kmer_content

-K

double

--max_mismatch_pvalue

-V

double

--fragment_length

-F

integer

--max_reads

-U

integer

--quantile

-Q

double

--exonic_fraction

-e

double

--top_n

-T

integer

--covered_fraction

-C

double

--max_itd_length

-l

integer

--min_itd_allele_fraction

-z

double

--min_itd_supporting_reads

-Z

integer

--skip_duplicate_marking

-u

boolean_true

--extra_information

-X

boolean_true

--fill_gaps

-I

boolean_true