HitSV: Advanced Structural Variant Detection Across Sequencing Technologies
A novel approach for precise genomic structural variant reconstruction using repetitiveness-aware assembly.
1Genomic Diversity Challenges
Structural variants are essential drivers of genomic diversity and disease susceptibility, yet their identification remains difficult. Repetitive genomic contexts and alignment ambiguities frequently hinder accurate detection. Current methods often struggle with the trade-off between sequencing costs and read lengths, leading to fragmented or misrepresented variants in complex, unstable genomic regions.
2Introducing HitSV Methodology
HitSV is a new alignment-based approach designed to function across long-read, short-read, and hybrid sequencing data. It explicitly models the interaction between local genomic repetitiveness and signature intensity. By redesigning core stages like site recognition and signature clustering, it effectively mitigates systematic distortions that typically occur in complex genomic regions.
3Breakpoint Recognition Strategy
The tool employs a repetitiveness-aware strategy that models structural variant signature densities using population-scale reference data. By constructing empirical distributions over various repeat contexts, it establishes data-type-specific criteria. This allows for sensitive detection in repeat-rich areas where traditional signals are often attenuated, obscured, or otherwise difficult to resolve accurately.
4Progressive Signature Clustering
HitSV utilizes a progressive clustering strategy to handle confounded breakpoint signals, which are common in large or repeat-associated rearrangements. By treating breakpoints as latent variables and iteratively refining their probability distributions, the tool aggregates fragmented evidence. This process disentangles overlapping signatures from distinct alleles, enabling the recovery of complex events.
5Haplotype-Resolved Local Assembly
The system performs haplotype-resolved local assembly using a unitig-centered de Bruijn graph that integrates both variant and nonvariant reads. For short-read data, coverage-informed traversal generates contigs beyond intrinsic limits. For long-read data, it clusters unitig-dBG paths, while hybrid data uses long-read contigs as backbones polished by short-read consensus sequences.
6Benchmarking Against SOTA
Performance was evaluated using 102 human datasets and 6 nonhuman species, comparing HitSV against 13 state-of-the-art callers. In the HG002 benchmark, HitSV achieved genotype F1 scores of 85.00% on PacBio HiFi and 88.48% on ONT data. These results outperformed existing callers by margins ranging from 15.34% to 71.68% across platforms.
7Cross-Species Performance Gains
HitSV demonstrated significant cross-species transferability, maintaining high performance across diverse animal and plant genomes. In multispecies benchmarks, it achieved average F1-score advantages of 4.83% to 14.54% for PacBio and 3.25% to 21.71% for ONT data. This consistency confirms the robustness of its empirical signature-distribution model across various biological contexts.
8Hybrid Sequencing Efficiency
The hybrid mode offers a cost-effective alternative to high-depth long-read sequencing for large cohorts. Using 4x long-read data combined with 30x short-read data, HitSV achieved performance comparable to or better than 30x long-read sequencing alone. It outperformed the hybrid caller blend-seq by 11.92% under matched sequencing depth and data conditions.
9Short-Read Sequencing Improvements
HitSV significantly enhances structural variant discovery from short-read data compared to single-reference callers like Manta and DELLY. In the HG002 benchmark, it achieved F1-GT scores 10.47% to 20.28% higher than these tools. The improvement remained consistent across both easy and hard genomic regions, demonstrating superior precision over integrated pipelines like GATK-SV.
10Reshaping Allele-Frequency Landscapes
When applied to the 1KGP phase 4 cohort, HitSV expanded the structural variant callset by 31.5%. This expansion substantially reshaped allele-frequency landscapes, providing a more comprehensive view of human genomic variation. The tool's ability to maximize discovery from existing datasets highlights its utility in large-scale population genomics and medical research studies.
11Uncovering Complex Architectures
In a large Chinese long-read cohort, HitSV uncovered tandem-repeat-mobile-element composite arrays as a prevalent class of complex structural variants. This finding underscores composite repeat organization as a previously underrecognized contributor to human structural variation. The tool's precision in repeat-rich regions allows for the identification of these complex, multi-allelic structural features.
12Impact on Genomic Discovery
By resolving previously inaccessible structural variants, HitSV provides a foundational improvement for population and medical genomics. Its ability to integrate diverse sequencing technologies while maintaining high precision enables more accurate variant reconstruction. As genomic studies scale, such robust, context-aware tools become essential for characterizing the full spectrum of human genetic variation.