Accurate Detection of Fusion Genes in Long-Read Transcriptome Datasets from Multiple Cancer Cell Lines

Fusion genes are observed in many cancer types and are suspected to have oncogenic properties. Long-read RNA-seq can sequence the full length of mRNA and facilitate detection of fusion genes. Several methods have been proposed for detecting fusion genes in long-read RNA-seq datasets extracted from cancer cells. However, the accuracy of the fusion-gene detection using those methods is reduced due to the relatively high percentage of false positives in the results. To cope with the problem, we have attempted to improve the detection accuracy by introducing several additional steps, anchoring breakpoints to exon boundary, gap re-alignment, and breakpoint clustering, in our fusion detection method. The detection performance is demonstrated by applying the method to fusion gene detection in the long-read RNA-seq datasets obtained from three cancer cell lines.