Performance of JAFFA and four other tools for different read lengths and layouts. We compared the performance of JAFFA, FusionCatcher, SOAPfuse, deFuse and TopHat-Fusion on the ENCODE dataset of the MCF-7 cell line, trimmed to emulate four different read configurations: single-end 50 bp (80 million reads), paired-end 50 bp (40 million read pairs), single-end 100 bp (40 million reads) and paired-end 100 bp (20 million read pairs). In each case, the total number of bases sequence was 4 billion. Only JAFFA and TopHat-Fusion could process single-end data. (A) Most true positives were reported with JAFFA on 100 bp paired-end reads followed by deFuse on 50 bp paired-end reads. (B) For each tool we compared the ranking of fusions, by selecting the read length and layout that maximised ROC performance. We ranked the candidates using the tools own scoring system, or if absent, the supporting data that maximised the area under the curve. JAFFA on 100 bp reads ranked true positives higher than any other combination.