Improving RNA Fusion Call Confidence and Reliability in Molecular Diagnostic Testing

Abstract

Next-generation sequencing is a superior method for detecting known and novel RNA fusions in formalin-fixed, paraffin-embedded tissue over fluorescence in situ hybridization and RT-PCR. However, confidence in fusion calling and true negatives may be compromised by poor RNA quality. Using a commercial panel of 507 genes and the recommended 3-million read threshold to accept results, two cases yielded false negatives while exceeding this recommendation during clinical validation. To develop a reliable quality control metric that better reflects internal sample quality and improves call confidence, gene expression across 361 patient tumor samples was evaluated to derive a set of 15 genes to serve as a proxy quality control (pQC). These 15 genes were assessed for their normalized expression using the sequencing data from each case and selected for robustness. A threshold of 11 pQC genes produced a 4.71% fail rate, selected for stringency as an acceptable level of repeated testing in the clinical setting, minimizing false-negative calls. To increase the chance that low-quality samples pass pQC, a revision to the library preparation method was also tested, with 75% of previously failed samples passing pQC on resequencing by increasing cDNA input. Taken together, a next-generation sequencing analysis quality control tool is presented that serves as a surrogate for housekeeping genes and improves confidence in fusion calls.