Fig 1.
Example of fusion detection from RNA-seq data.
The fusion consists of exons 1 and 2 from gene A and exons 3 and 4 from gene B. SPLIT reads cover the junction point, while BRIDGE reads span the junction point within the insert region, which is not sequenced.
Fig 2.
Outline of the different steps of the analysis pipeline. Properly aligned pairs require the distance between alignments to be within the maximum intron size on the same chromosome.
Fig 3.
Clustering of breakpoint candidates.
The arrows of the SPLIT alignments and the dot lines of BRIDGE alignments demonstrate the direction to the breakpoint position. (A) Initial clusters are created from intersecting SPLIT and BRIDGE alignments. (B) Cluster 4 is separated from cluster 1 based on the directionality, which is inferred from the alignment strand and order. (C) Cluster 5 is separated from cluster 2 based on the putative breakpoint position. Alignments belonging to the same breakpoint candidate have the same color. BRIDGE reads are marked with b, SPLIT reads are marked with s. A SPLIT read assumes an exact breakpoint, while a BRIDGE read assumes an approximate breakpoint within allowed insert size distance.
Table 1.
Comparison of fusion detection tools on simulated data.
Fig 4.
Comparison of recall and precision on simulated data.
Recall and precision are plotted based on the number of supporting reads. For each given threshold we selected fusions that by simulation design have a number of supporting reads less or equal to the threshold. The number of true positive events was computed for every tool using only the selected fusions.
Table 2.
Fusion events detected in public RNA-seq datasets.
Table 3.
RNA-Seq sample details.
Table 4.
Fusions selected for qRT-PCR validation from RNA-seq of VCaP and LNCaP cell lines.
Fig 5.
(A) Genomic structure of the TMPRSS2–ERG fusion transcripts discovered from deep sequencing data by InFusion. Isoform 3 is a known transcript, while isoforms 1 and 2 are novel. Transcript names are taken from the Ensembl v.68 database. (B) RT-PCR validation of isoforms in VCaP, LNCaP, RWPE-1 and PrEC cell lines; NTC = no template control. The PCR primer design was based on the output from the InFusion pipeline. In order to detect only one product, one PCR primer specific for Isoform 3 was designed to cover the fusion junction site. A 50 bp DNA ladder was co-run as size marker; bright bands indicate 250 bp and 500 bp. (C) Relative expression levels of the fusion isoforms as measured by qRT-PCR. All measurements were performed in triplicate, mean expression values were computed relative to GAPDH. Plotted values are normalized to the computed expression of isoform 3. (D) Expression levels of isoforms estimated in RPKM under the assumption of uniform coverage.