Exogene: A performant workflow for detecting viral integrations from paired-end next-generation sequencing data

Abstract

The integration of viruses into the human genome is known to be associated with tumorigenesis in many cancers, but the accurate detection of integration breakpoints from short read sequencing data is made difficult by human-viral homologies, viral genome heterogeneity, coverage limitations, and other factors. To address this, we present Exogene, a sensitive and efficient workflow for detecting viral integrations from paired-end next generation sequencing data. Exogene's read filtering and breakpoint detection strategies yield integration coordinates that are highly concordant with long read validation. We demonstrate this concordance across 6 TCGA Hepatocellular carcinoma (HCC) tumor samples, identifying integrations of hepatitis B virus that are also supported by long reads. Additionally, we applied Exogene to targeted capture data from 426 previously studied HCC samples, achieving 98.9% concordance with existing methods and identifying 238 high-confidence integrations that were not previously reported. Exogene is applicable to multiple types of paired-end sequence data, including genome, exome, RNA-Seq and targeted capture.

Fig 1. Overview of Exogene workflow.

Fig 2. Comparison of evidence for HBV integration at chr7:72,027,703 from Exogene and HGT-ID.

Shaded regions indicate breakpoint ranges as inferred from read fragment lengths and orientations, darker shades indicate greater support.

Fig 3. Concordance rate of Exogene and HIVID calls as a function of minimum coverage.

Fig 4. Concordance rate of Exogene and HIVID calls as a function of minimum allowable percentage of reads aligned with mapping quality 0.