Accurate and efficient detection of gene fusions from RNA sequencing data.
Uhrig S, Ellermann J, Walther T, Burkhardt P, Fröhlich M, Hutter B, Toprak UH, Neumann O, Stenzinger A, Scholl C, Fröhling S, Brors B. · 2021
Read Accurate and efficient detection of gene fusions from RNA sequencing data. by Uhrig S, Ellermann J, Walther T, Burkhardt P, Fröhlich M, Hutter B, Toprak UH, Neumann O, Stenzinger A, Scholl C, Fröhling S, Brors B. (2021) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/europmc-pmc7919457.txt.
Introduction
Uhrig and co-authors present a research study on finding gene fusions in RNA sequencing data. Fusions arise when two genes join into one abnormal message. They drive many cancers and can be hit by targeted drugs. Finding them in sequence data is a routine task, yet existing tools are often slow or inaccurate.
The team built a new algorithm called Arriba. It pairs high sensitivity with a short runtime. On 803 published pancreatic cancer samples, Arriba found many driver fusions. Several hit proteins that drugs can target, such as ALK and FGFR2. The fusions clustered in tumours without KRAS mutations, as if standing in for them. Two new fusions changed cells in lab tests, proving they can drive growth. The authors offer the tool for both research and clinical use.
Catalog record
| CREATOR | Uhrig S, Ellermann J, Walther T, Burkhardt P, Fröhlich M, Hutter B, Toprak UH, Neumann O, Stenzinger A, Scholl C, Fröhling S, Brors B. |
|---|---|
| YEAR | 2021 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 75 KB |
| LICENSE | CC BY 4.0 |
| RIGHTS | CC BY |
| SOURCE | Europe PMC |
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
Accurate and efficient detection of gene fusions from RNA sequencing data By Sebastian Uhrig, Julia Ellermann, Tatjana Walther, Pauline Burkhardt, Martina Fröhlich, Barbara Hutter, Umut H Toprak, Olaf Neumann, Albrecht Stenzinger, Claudia Scholl, Stefan Fröhling, Benedikt Brors Published in: Genome Research DOI: 10.1101/gr.257246.119 ABSTRACT Abstract The identification of gene fusions from RNA sequencing data is a routine task in cancer research and precision oncology. However, despite the availability of many computational tools, fusion detection remains challenging. Existing methods suffer from poor prediction accuracy and are computationally demanding. We developed Arriba, a novel fusion detection algorithm with high sensitivity and short runtime. When applied to a large collection of published pancreatic cancer samples (n = 803), Arriba identified a variety of driver fusions, many of which affected druggable proteins, including ALK, BRAF, FGFR2, NRG1, NTRK1, NTRK3, RET, and ROS1. The fusions were significantly associated with KRAS wild-type tumors and involved proteins stimulating the MAPK signaling pathway, suggesting that they substitute for activating mutations in KRAS. In addition, we confirmed the transforming potential of two novel fusions, RRBP1-RAF1 and RASGRP1-ATP1A1, in cellular assays. These results show Arriba's utility in both basic cancer research and clinical translation. Gene fusions play a major role as oncogenic drivers in many cancer types. This insight has immediate consequences for the treatment of patients, because many gene fusions can be addressed therapeutically with targeted drugs (Schram et al. 2017). The most prominent examples are fusions between BCR and ABL1 in chronic myeloid leukemia and acute lymphoblastic leukemia, which can be treated effectively using imatinib and related drugs (An et al. 2010). More recently, the U.S. Food and Drug Administration (FDA) granted accelerated approval for the treatment of solid tumors harboring NTRK fusions with larotrectinib after showing antitumor activity in three multicenter trials (Drilon et al. 2018). For this reason, a routine task in genomics-guided precision oncology is to search for evidence of gene fusions in RNA sequencing (RNA-seq) data. Although a variety of computational tools for the detection of gene fusions have been developed over the years, there is still no gold standard. The reliable prediction of gene fusions from short-read RNA-seq has proven to be difficult owing to a myriad of artifacts being introduced during library preparation and sequence alignment. To keep the number of false-positive predictions low, the algorithms implement stringent filters, with the undesired side effect that occasionally driver fusions are discarded and events with subtle evidence in RNA-seq data are lost entirely. The current practice is to apply at least two tools and use the union or intersection of their predictions. This approach is computationally expensive, because each tool on its own typically takes many hours or even days to run. With high-throughput sequencing (HTS) technology becoming more common in clinical practice to identify targetable alterations, the demand for algorithms that are both accurate and efficient grows. Inaccurate predictions complicate the interpretation of HTS-based results, and the time-critical operation of a precision oncology trial does not tolerate slow computational pipelines, because the overall workflow allocates only a few days for bioinformatics processing (Roychowdhury et al. 2011; Worst et al. 2016). We developed Arriba, a fusion detection algorithm specifically designed to meet the demanding requirements of HTS-assisted precision oncology. Owing to a highly optimized workflow, it can process contemporary RNA-seq samples in less than an hour. Sophisticated filters detect fusions even under unfavorable conditions, such as low sample purity. In addition, Arriba is capable of detecting aberrant transcripts that are not called by most fusion detection methods but may be clinically relevant. This includes tumor suppressor genes that are occasionally inactivated by rearrangements within the gene or by translocations to introns or intergenic regions. Even though the same technical approach can be applied to detect such transcripts, most available fusion detection tools do not report them. As a consequence, clinically relevant aberrations may be overlooked. Arriba improves over existing methods in that it can find intragenic inversions/duplications and translocations to introns/intergenic regions. Based on novel gene fusions involving NRG1 that we recently discovered in a series of patients with KRAS wild-type pancreatic tumors (Heining et al. 2018) in the context of NCT/DKTK MASTER (Horak et al. 2017), a HTS-guided precision oncology program, we applied Arriba to further explore the relevance of gene fusions in pancreatic cancer. In particular, we investigated the prevalence of druggable fusions, because there are few targeted treatment options for pancreatic cancer patients, and despite recent improvements in conventional and targeted therapies (Conroy et al. 2018; Golan et al. 2019), the 5-yr overall survival rate is <10%. RESULTS We compared the performance of Arriba v1.0.0 against six commonly used fusion detection algorithms (defuse v0.8.1 [McPherson et al. 2011], FusionCatcher v1.00 [Nicorici et al. 2014], InFusion v0.8 [Okonechnikov et al. 2016], PRADA v1.2 [Torres-García et al. 2014], SOAPfuse v1.27 [Jia et al. 2013], STAR-Fusion v1.4.0 [Haas et al. 2019]) with respect to speed and accuracy. Accuracy benchmarks To show Arriba's robust performance across diverse types of input data, we assessed its accuracy on four types of peer-reviewed benchmark data sets (Supplemental Tables S1, S2): We used in silico–generated fusion transcripts from Jia et al. (2013), who simulated 150 fusion transcripts and merged them into an RNA-seq sample from benign tissue (H1 human embryonic stem cells), serving as background expression. To measure the sensitivity of a method as a function of the expression level of a fusion transcript, nine different expression levels were simulated, ranging from five- to 200-fold. Next, we took RNA-seq samples from Tembe et al. (2014), who used a semisynthetic approach to benchmark fusion detection algorithms. The investigators spiked in synthetic RNA molecules into RNA libraries of the melanoma cell line COLO-829. The synthetic RNA molecules mimic the transcript sequences of nine oncogenic fusions found in a variety of cancer types. They were spiked into 20 replicates of RNA libraries at 10 different concentrations ranging from 10−8.57 pMol to 10−3.47 pMol. In addition, one endogenous fusion of the COLO-829 cell line was confirmed via orthogonal validation. To measure the performance on real data, we ran the tools on eight samples from four cell lines (Edgren et al. 2011; The ENCODE Project Consortium 2012; Lin et al. 2015), including the breast cancer cell line MCF-7, a well-studied cancer cell line with a highly rearranged genome and many gene fusions validated via orthogonal methods. We used the list of validated fusions compiled by Davidson et al. (2015), which comprises 69 distinct pairs of fusion genes. Because this list is biased toward fusions that were detected by previous methods and potentially lacks events that can be detected by newer, more sensitive methods, we also considered a prediction to be true if its breakpoints were close to the breakpoints of a structural variant identified in a whole-genome sequencing (WGS) sample of the MCF-7 cell line (Li et al. 2016). Furthermore, we subjected the top predictions of each tool to experimental validation if they were confirmed neither by previous validation tests nor by structural variants (Supplemental Table S3). Lastly, we applied the tools to patient data from the ICGC early-onset prostate cancer cohort (ICGC-EOPC) and the TCGA diffuse large B-cell lymphoma cohort (TCGA-DLBC). Early-onset prostate cancer is characterized by a high prevalence of TMPRSS2-ERG fusions