EchoSV: diploid donor-specific assembly enhances somatic structural variant detection in cancer genomes

"echosv_test_data.tar.gz" contains a small dataset designed for a quick-start demonstration and testing of the EchoSV pipeline. "echosv_results.tar.gz" contains the full pipeline output for generating multi-platform somatic structural variant (SV) high-confidence sets across six cancer cell lines. It includes a comprehensive comparison across multiple references using the EchoSV toolkit. The data is structured to ensure transparency and ease of use: De Novo Assemblies: The six dual-haplotype donor-specific assemblies (DSAs) generated with hifiasm. Raw SV Callsets: Original SV calls produced by each calling strategy for every sample and reference. High-Confidence Somatic SV Callset: Filtered somatic SVs defined by the support criteria (≥2 of 3 sequencing platforms and ≥4 of 15 callers). Scripts for reproducing the SV merging and comparison workflow are also included (01_sv_merge_15callset.sh; 02_sv_consolidate_dsa.sh; 03_sv_compare_refs.sh) Final Truth Set Annotation: Each high-confidence SV is annotated with its assembly-based validation status, overlap with satellite regions, and cross-reference status. Events rejected during this evaluation are tagged with GermlineError or MapError. The data follows the EchoSV processing flow:raw_calls/ (9 callers × 4 refs × 6 samples) ↓ echosv merge high_confidence_calls/ (*_highconfi.vcf.gz) ↓ echosv genotype (LR + SR BAMs) ↓ echosv match (cross-haplotype / cross-reference)final_truthset/ (*_truthset.vcf.gz)