10. Standardizing fusion calls in a computable format with FUSOR for downstream clinical assessment

Abstract

The detection of gene fusion events, in which two or more genes interact to drive aberrant expression of a gene product, plays a key role in clinical diagnostics. Although advances in sequencing technology have strengthened gene fusion data availability, there are limitations in the way such knowledge is interpreted in a clinical context. Specifically, current standards are imprecise for representing the complexity of fusions that are observed from biological specimens. To address this challenge, experts from VICC, CGC, and other clinical genomics communities developed a consensus, unified framework for the description of fusion events. We developed the FUSOR package, a Python library containing modeling and validation tools that implements this standard for use with gene fusion data.

We tested use of FUSOR on patient sample data through development of a Translator module (http://tinyurl.com/FUSOR-Translator) that standardizes fusion calls from eight widely-used fusion detection algorithms including CICERO and Arriba and fusion calls from the AACR Project GENIE cohort. We assessed application of the VICC Gene Fusion Specification using FUSOR to evaluate the completeness of the specification for representing fusion variant calls. We demonstrate how application of the tool to real-world data identified gaps in the nascent specification, including the use of gene concepts not covered by the HUGO Gene Nomenclature committee and the improved alignment of evidence between assayed and categorical fusion concepts, that we were able to fill to improve the standard. We conclude with applications of the FUSOR tool for use with clinical variant curation workflows.