Somatic Structural Variant Discovery from Tumor Specific Strings Computed Against a Draft Composite Germline Reference

Abstract

Somatic structural variants (SVs) are among the most clinically relevant genomic alterations in cancer. However, their accurate detection remains challenging due to complex genomic rearrangements, repetitive genomic regions, difficulties in distinguishing somatic variants from germline variation and sequencing artifacts, and limited sequencing support for low-frequency events. Existing mapping-based somatic SV callers are susceptible to alignment ambiguities in repetitive regions and near complex SV breakpoints, while approaches relying on comparisons of independently generated tumor and normal variant callsets may propagate errors from both callsets. Building on the concept of specific strings, previously applied to germline SV discovery, we introduce VESPA-SS, a novel approach for detecting somatic SVs from paired tumor-normal long-read whole-genome sequencing data. VESPA-SS directly extracts somatic SV signals using specific strings computed from tumor reads against a Draft Composite Germline Reference, without requiring initial read mapping or a high-quality germline assembly. These signals are subsequently clustered and refined through consensus sequence generation and alignment-based procedures to identify somatic SVs. Evaluation on real PacBio HiFi datasets demonstrates that VESPA-SS outperforms state-of-the-art somatic SV callers, achieving F1 scores exceeding 0.92 and 0.90 on the well-characterized HG008 and COLO829 cancer cell lines, respectively, compared with the highest F1 scores of 0.86 and 0.85 achieved by the evaluated methods. These results demonstrate the effectiveness of specific string based signal extraction for accurate somatic SV discovery and establish a new computational paradigm for identifying complex genomic alterations in cancer. VESPA-SS is freely available at https://github.com/dcmonti/VESPA.