Highly Sensitive and Specific Method for Detection of Clinically Relevant Fusion Genes across Cancer

Gene fusions are strong driver mutations in cancer and can be used as a diagnostic tool to predict different tumour phenotypes and treatments. Several fusion detection algorithms for RNA-Seq data have been developed, but all of them report a consistently high number of false positive events. Therefore, new methods are crucial to accurately identify potential fusions that may be key drivers of oncogenesis. We developed Fusion Validator, a new filtering tool able to discriminate false positive fusion transcripts from real fusions and significantly reduce the number of candidates to assess for experimental validation. Fusion Validator perform a local realignment of reads on each fusion transcript sequence and tries to close the gap around the fusion breakpoint using both a de novo assembly and a seed-extend algorithm. If the algorithm fails to reconstruct the fusion transcript around the breakpoint, the fusion is considered as false positive and is discarded. Additional filtering steps are used to remove fusions with breakpoints mapping on low complexity or homologous regions and to find correct fusion partners for promiscuous gene fusion events. A final ranking score based on fusion annotation is created for each validated event to help distinguish real driver fusions from passengers one. We tested Fusion Validator on simulated datasets of different coverage, read length and breakpoint positions, and on four published breast cancer Cell Lines, highlighting the massive increase in sensisitivity, precision and specificity of our algorithm, in comparison to other fusion-detection software. Using this tool, we successfully detected 97.95% of PCR-validated kinase recurrent fusions in 190 pan cancer samples, removing approximately 79.95% of false positives. Particularly in haematological disorders and childhood sarcomas, gene fusions are critical as diagnostic and prognostic factors. Therefore, development of this novel tool to increase the efficiency of detecting driver fusions is critical in disease detection and treatment