Detecting structural variants in AML using Oxford Nanopore long-reads

Abstract

Structural variants (SVs) are, like single nucleotide variants and indels, a major driving force of cancer development. Accurate detection of pathogenic SVs in patients enables targeting of the underlying molecular cause of the cancer, which leads to possibilities of a specialized treatment plan and in turn a better prognosis. However, traditional molecular diagnostic methods require prior knowledge of the SV location and have a low throughput compared to next-generation sequencing methods. Calling SVs using (Illumina) short-read sequencing is currently the standard approach, but long repetitive regions are a major problem for SV calling based on short-read sequencing technologies. Long-read sequencing technologies (e.g. ONT) are potentially more suitable for SV detection in these regions, but have a lower single-nucleotide accuracy, are more expensive and are less mature platforms. In this report, we propose a workflow for SV calling using ONT long-reads. Additionally, we benchmark the workflow and show results from three pediatric AML samples to show the efficacy on real-world data.