Rapid and Infinitely Scalable Fusion Gene Detection in the Cloud

S. Newman,Y. Li,X. Zhou,C. McLeod,M. Rusch,J. Easton,S. V. Rice,S. A. Shurtleff,J. Nakitandwe,E. M. Azzato,K. E. Nichols,J. R. Downing,D. W. Ellison,J. Zhang

DOI: https://doi.org/10.1016/j.cancergen.2017.04.007

IF: 2.169

2017-01-01

Cancer Genetics

Abstract:Pediatric tumors undergo a battery of time-sensitive molecular and cytogenetic tests to detect gene fusions and other structural variants. While many labs are beginning to use targeted NGS panels to detect such events, we considered whole transcriptome sequencing (RNA-Seq) to be the ultimate solution as it samples all expressed loci including those only relevant to rare pediatric tumors. Clinical RNA-Seq, although an attractive option, presents several unique challenges including i) How to avoid an informatics bottleneck within a time-sensitive workflow ii) How to minimally burden those reviewing the results and iii) How to standardize and share complex computational analysis methods. Here we present a pilot study of clinical RNA-Seq analyzed in the cloud. Using a highly secure platform, we developed our Rapid RNA-Seq pipeline centered on an extensively validated fusion gene detection algorithm “CICERO.” We initially detected fusion transcripts and structural variants in 78 diverse but well-characterized samples and showed that Rapid RNA-Seq detected all abnormalities found previously. We next ran Rapid RNA-Seq in parallel with standard testing in real time, again showing concordance, but also detecting a large cache of previously undetected events—some of which were targetable. Run times averaged approximately four hours from raw data acquisition to reporting and were not affected by caseload. To aid in clinical review, we developed a rich graphical interface based on our ProteinPaint software (https://pecan.stjude.org/#/home) that allowed pathologists to easily assess the supporting evidence and functional impact of any given event. We make our tools and workflows securely available at www.stjude.cloud.