Genomic structural variation and associated RNA fusions are a common clinical feature known to be involved in the initiation and pathogenesis of cancer. This complex class of variants also has significant implications on therapeutic decisions and has emerging roles in evidence-based clinical applications. Consequently, fusion detection is an emerging aspect of precision medicine that enables clinical utility of fusions as a therapeutically-relevant target which includes small molecules, biologics, and neoantigen-directed approaches.Here we present STAR-SEQR, an algorithm that uses multiple levels of evidence to detect and quantify fusions from RNA-Seq data. STAR-SEQR is a fast and accurate tool that goes beyond fusion detection and also provides rich annotation and useful reporting features to aid in the adoption of fusions in clinical diagnostics. Notably the software produces fusion expression values, PCR primers, fusion RNA sequence, and predicted peptide sequence to facilitate downstream applications.To address the analytical performance of STAR-SEQR to existing tools, we utilized previously characterized synthetic datasets and real-world sample data with matching DNA sequencing. Of note, in a dataset of 50 synthetic fusions, STAR-SEQR had the highest accuracy of all tools tested and called 49/50 events with no false positives. STAR-SEQR has also been run on several large clinical datasets and the fusion events identified have displayed strong overlap with identified DNA structural variants. STAR-SEQR also performed well in the ongoing ICGC-TCGA DREAM SMC-RNA Challenge and achieved the highest F1 score in the simulated data round with an overall average of 0.96. Analytical lab testing has also been performed using the Seracare Fusion RNA Mix V2 sample containing 15 clinically significant fusions with known concentrations in the low range of detectability. STAR-SEQR accurately called 13/15 events with the other events having no identifiable read alignments. We have also benchmarked methods comparing FFPE material, range of read lengths, and library methods where STAR-SEQR demonstrated excellent tradeoff between sensitivity and precision and generally outperformed other tools.In summary, STAR-SEQR has excellent analytical performance that surpasses existing fusion detection tools. It is broadly applicable to RNA-Seq methods, computationally efficient, and robust across sample characteristics. Lastly, STAR-SEQR is explicitly designed to enable fusion-directed clinical diagnostics and support cancer immunotherapy approaches that utilize fusion neoantigens.

Citation Format: Jeff Jasper, Jason G. Powers, Victor J. Weigman. STAR-SEQR: Accurate fusion detection and support for fusion neoantigen applications [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2018; 2018 Apr 14-18; Chicago, IL. Philadelphia (PA): AACR; Cancer Res 2018;78(13 Suppl):Abstract nr 2296.