JAFFA: High sensitivity transcriptome-focused fusion gene detection

Table 1 A comparison of fusion detection performance on simulated RNA-Seq, 75 bp simulation of 50 fusions

Tool	True positives	Sensitivity	False positives
JAFFA - Hybrid	44 (32/12/0)	88%	0
JAFFA - Assembly	39 (28/11/0)	78%	0
SOAPfuse	37	74%	1
JAFFA - Direct	34 (32/2/0)	68%	0
deFuse	34	68%	0
TopHat-Fusion	27	54%	0
FusionCatcher	Unable to run on a low number of reads

We ran all three modes of JAFFA in addition to SOAPfuse, TopHat-Fusion, deFuse and FusionCatcher on a simulation set of 57,000 75 bp RNA-Seq read pairs provided with FusionMap. JAFFA had the highest sensitivity when run in Hybrid mode, identifying 44 out of 50 possible fusion events. For all JAFFA modes, no false positives were reported. In parenthesis we show the value at each of JAFFA’s classification levels: (high/medium/low) confidence.

ISSN: 1756-994X