Logo of bioinfoLink to Publisher's site
Bioinformatics. Jul 15, 2012; 28(14): 1923–1924.
Published online May 4, 2012. doi:  10.1093/bioinformatics/bts272
PMCID: PMC3389765

BreakFusion: targeted assembly-based identification of gene fusions in whole transcriptome paired-end sequencing data


Summary: Despite recent progress, computational tools that identify gene fusions from next-generation whole transcriptome sequencing data are often limited in accuracy and scalability. Here, we present a software package, BreakFusion that combines the strength of reference alignment followed by read-pair analysis and de novo assembly to achieve a good balance in sensitivity, specificity and computational efficiency.

Availability: http://bioinformatics.mdanderson.org/main/BreakFusion

Contact: gro.nosrednadm@3nehck; ude.ltsuw.emoneg@gnidl

Supplementary information: Supplementary data are available at Bioinformatics online


Despite recent progress, bioinformatics tools that analyze RNA-seq data are often limited in accuracy and scalability. Alignment-based tools (Garber et al., 2011) typically predict such large numbers of candidates that further examination of the data is impractical. Furthermore, many findings are artifacts derived from a large number of erroneous short reads produced by the next-generation sequencing (NGS) instruments or are read misalignments produced by the sequence aligners. Assembly-based approaches (Martin and Wang, 2011) can potentially achieve higher accuracy because they leverage dependency among reads and are less sensitive to errors in individual reads. With sufficient coverage, longer and higher quality sequences can be assembled from short reads, resulting in improved specificity. However, existing assembly-based approaches (Garber et al., 2011; Grabherr et al., 2011; Robertson et al., 2010; Trapnell et al., 2010) potentially lack sensitivity to rare events such as gene fusions because they are designed to reconstruct entire transcriptomes, rather than focus on novel sequences. We reason that a targeted transcriptome assembly strategy that focuses on assembling novel junction sequences can potentially achieve a better balance in sensitivity, specificity and computational efficiency.


Here we present an NGS Bioinformatics pipeline, BreakFusion that implements a targeted assembly approach for novel transcriptomic sequence discovery.

BreakFusion consists of five steps, as illustrated in Figure 1. The input is one or a set of whole transcriptome BAM files that contains mapped paired-end RNA-seq reads. The mapping can be performed using a set of algorithms such as BWA and TopHat, preferably including known splice junctions in the references to achieve good alignment. Step 1 identifies splicing breakpoints using a read-pair algorithm such as BreakDancer (Chen et al., 2009) or any splice mapping algorithm such as TopHat-Fusion (Kim and Salzberg, 2011). Step 2 locally assembles short reads that are anchored around each breakpoint using TIGRA and produces a set of splice junction contigs that are supported by both mapped and one-end-anchored reads (Mills et al., 2011). TIGRA is a de Bruijn graph-based assembler that is specially designed to construct all possible allelic sequences at loci of heterogeneous origin. It is sufficiently sensitive to identify low abundance alleles as long as their constituent kmers are observed at least twice in the reads. Step 3 aligns the junction sequences to the genome using BLAT (Kent, 2002). Step 4 summarizes BLAT alignments into a single chimeric score that quantifies the likelihood of an assembled junction sequence containing bona fide breakpoints relative to the genome. Step 5 annotates the breakpoints using UCSC databases and filters the breakpoints by mainly two factors: (i) the chimeric scores computed in Step 4, and (ii) the self-chain alignment annotation, which indicates whether the breakpoints are caused by misalignment in duplicated regions.

Fig. 1.
BreakFusion overview

Most steps in the BreakFusion pipeline are performed by previously published and well-attested algorithms. The chimeric scoring system in Step 4 is novel to this work. By default, BLAT does not produce a single hit for transcripts that span two chromosomes or distances >750 000 bp. Instead, it reports individual alignments for each of the sub-segments, with no direct indication of the existence of the chimeric structure and the level of confidence associated with it. To overcome such limitation, we first remove hits that are not unique (i.e. map to multiple regions). We then chain the remaining alignments into longer ones if they can form 1-monotonic maps (Brudno et al., 2003). We compute scores for chained alignment using the same BLAT formula that is used for each constituent alignment. We identify the best and the second best alignment after chaining, and compute a chimeric score using the following equation:

equation image

Where q* is the alignment score of the best chained alignment, q′ the second best, and q0 is the length of the query sequence. This equation produces scores between 0 and 1.0. The score becomes high when the best alignment well explains the entire query sequence (q* approaches q0) and the second best alignment is appreciably worse (q[double less-than sign]q*). The constant 10 makes the scores sensitive to differences at the 10-bp scale.


To test BreakFusion's performance, we compared it with TopHat-Fusion and defuse (McPherson et al., 2011) using four publicly available RNA-seq datasets: a prostate cancer cell line (NCIH660) and a matched lymphoblastoid cell line (GM12878) (Sboner et al., 2010), a breast cancer cell line (MCF-7) (Edgren et al., 2011) and a chronic myelogenous leukemia cell line (K-562) (Berger et al., 2010). BreakFusion achieved better sensitivity and specificity tradeoff in these tests (Supplementary Data). It also used substantially less CPU time. We further applied BreakFusion to analyze RNA-seq data for 155 acute myeloid leukemia (AML) samples as part of The Cancer Genome Atlas (TCGA) project. This dataset was also analyzed by the Genome Sciences Centre at the BC Cancer Agency (BCCA) using Tran-Abyss (Robertson et al., 2010), which resulted in 67 instances of experimentally validated fusions. BreakFusion was able to rediscover all these fusions and additionally predicted eight new instances of fusions, which were subsequently confirmed by the BCCA group (Supplementary Data).


To the best of our knowledge, BreakFusion is the first approach that performs targeted assembly on RNA-seq data for fusion identification. The results of our experiments indicate that BreakFusion has achieved sensitivity and specificity comparable or better than other tools, and is clearly more computationally efficient. This, in our view, represents a methodology improvement that will benefit many projects that utilize NGS RNA sequencing data. Besides fusion discovery, in principle, BreakFusion can be applied to identify novel alternative splicing events.

The component programs of BreakFusion are efficiently implemented in C++ and perl. It finished analyzing 155 TCGA AML BAM files in <2 h using 155 CPUs with 4 GB RAM each. BreakFusion is freely available for academic use at http://bioinformatics.mdanderson.org/main/BreakFusion.

Supplementary Material

Supplementary Data:


We thank Hao Zhao, Jianping Zhang, Heather Schmidt and Christopher Miller for their assistance. We are also grateful to Nina Thiessen, Elizabeth Chun and Gordon Robertson from the BCCA, as well as the members of TCGA AML Analysis Working Group.

Funding: The Cancer Center Support Grant from NCI (P30 CA016672 to UT MD Anderson); Center for Large-Scale Genome Sequencing and Analysis from NHGRI (U54 HG003079 to R.K.W); Structural Genomic Variation analysis for the 1000 Genome project from NHGRI (U01 HG005209 to L.C.)

Conflict of Interest: none declared.


  • Berger M.F., et al. Integrative analysis of the melanoma transcriptome. Genome Res. 2010;20:413–427. [PMC free article] [PubMed]
  • Brudno M., et al. Glocal alignment: finding rearrangements during alignment. Bioinformatics. 2003;19(Suppl. 1):i54–i62. [PubMed]
  • Chen K., et al. BreakDancer: an algorithm for high-resolution mapping of genomic structural variation. Nat Methods. 2009;6:677–681. [PMC free article] [PubMed]
  • Edgren H., et al. Identification of fusion genes in breast cancer by paired-end RNA-sequencing. Genome Biol. 2011;12:R6. [PMC free article] [PubMed]
  • Garber M., et al. Computational methods for transcriptome annotation and quantification using RNA-seq. Nat. Methods. 2011;8:469–477. [PubMed]
  • Grabherr M.G., et al. Full-length transcriptome assembly from RNA-Seq data without a reference genome. Nat. Biotechnol. 2011;29:644–652. [PMC free article] [PubMed]
  • Kent W.J. BLAT - The BLAST-like alignment tool. Genome Res. 2002;12:656–664. [PMC free article] [PubMed]
  • Kim D., Salzberg S.L. TopHat-Fusion: an algorithm for discovery of novel fusion transcripts. Genome Biol. 2011;12:R72. [PMC free article] [PubMed]
  • Martin J.A., Wang Z. Next-generation transcriptome assembly. Nat. Rev. Genet. 2011;12:671–682. [PubMed]
  • McPherson A., et al. deFuse: an algorithm for gene fusion discovery in tumor RNA-Seq data. PLoS Comput. Biol. 2011;7:e1001138. [PMC free article] [PubMed]
  • Mills R.E., et al. Mapping copy number variation by population-scale genome sequencing. Nature. 2011;470:59–65. [PMC free article] [PubMed]
  • Robertson G., et al. De novo assembly and analysis of RNA-seq data. Nat Methods. 2010;7:909–U962. [PubMed]
  • Sboner A., et al. FusionSeq: a modular framework for finding gene fusions by analyzing paired-end RNA-sequencing data. Genome Biol. 2010;11 [PMC free article] [PubMed]
  • Trapnell C., et al. Transcript assembly and quantification by RNA-Seq reveals unannotated transcripts and isoform switching during cell differentiation. Nat. Biotechnol. 2010;28:511–515. [PMC free article] [PubMed]

Articles from Bioinformatics are provided here courtesy of Oxford University Press
PubReader format: click here to try


Related citations in PubMed

See reviews...See all...

Cited by other articles in PMC

See all...


  • PubMed
    PubMed citations for these articles

Recent Activity

Your browsing activity is empty.

Activity recording is turned off.

Turn recording back on

See more...