Applications of NGS to Screen FFPE Tumours for Detecting Fusion Transcripts

Fusion transcripts play an important role in a variety of human cancers. But bioinformatics algorithms that use RNA-Seq data to detect fusions tend to yield high false-positive rates due to both relatively short reads from next-generation sequencing (NGS) and the repetitive elements in human genome. The primary purpose of this chapter is to discuss the design strategy of the bioinformatics methods used for detection of fusion transcripts and to compare their strengths and weaknesses, or “fit for purpose,” on RNA-Seq data from non-fixed or formalin-fixed paraffin-embedded (FFPE) tumor tissues. A large number of archival FFPE tumor tissue samples are associated with mature medical records including disease outcome; these samples offer great potential for diagnostic and therapeutic target discovery. However, the chemical treatment by formalin causes RNA degradation and base deamination, which lead to low library complexity and mapping quality. It is important that bioinformatics tools are designed to address these challenges. Here we illustrate a framework to address them, using gFuse as the example. We present results by comparing the fusion transcripts discovered from analysis of RNA from fresh and FFPE MCF-7 breast cancer cells. We also describe the application of gFuse to RNA-Seq data generated from two independent breast cancer cohorts with clinical outcomes and identify candidate fusion transcripts relevant to disease progression.