DNA shape and epigenomics distinguish the mechanistic origin of human genomic structural variations

Nadejda Boev, Mark Gerstein, Sushant Kumar

TL;DR

This paper introduces a novel computational framework that goes beyond simple sequence homology to decipher the underlying repair mechanisms responsible for human structural variations (SVs). By integrating DNA shape, epigenomic data, and an active-learning approach, the study reveals distinct features associated with different SV origins and their potential links to disease.

What Problem Does This Paper Address?

Structural variations (SVs), which are large-scale changes in DNA like insertions, deletions, and rearrangements, are significant drivers of human genetic diversity and disease. However, accurately identifying the mechanisms that cause these SVs has been challenging. Traditional methods often rely on local sequence homology, which can be insufficient to capture the full picture. Double-strand breaks (DSBs) are a primary source of SVs, and their repair pathways (like homologous recombination or non-homologous end joining) are influenced by a complex interplay of genomic and epigenomic factors. This paper aims to bridge this gap by developing a more comprehensive approach to infer these repair mechanisms, moving beyond simple sequence similarity.

Methods

The authors developed a multi-stage computational approach:

  1. Homology-Based Workflow: Initially, SVs were classified based on local sequence homology around their breakpoints. This established a baseline categorization into "High Local Homology" (HLH), "Intermediate Local Homology" (ILH), and "No Local Homology" (NLH) classes, inspired by known DSB repair pathways.
  2. Feature Annotation: SVs and their flanking regions (2000 bp upstream and downstream) were annotated with a wide array of features, including sequence characteristics (%GC, stability), DNA shape (minor groove width, helical twist), epigenomic profiles (histone modifications, DNA accessibility from ChIP-Seq and DNase-seq), replication timing, repetitive elements, and non-B DNA structures.
  3. Standardization and Z-scores: To compare features across different SVs and simulations, a z-score standardization was applied, comparing observed SV features to those from simulated SVs.
  4. Active-Learning Unsupervised Clustering: A key innovation is the use of an active-learning-based unsupervised clustering approach (HDBSCAN). This method leverages the standardized features to identify SV clusters that reflect underlying repair mechanisms, without strictly relying on homology thresholds. This allowed for the discovery of SVs with similar repair contexts even if their local homology differed.
  5. Sub-Clustering (HD/nHD): Within the identified clusters, SVs were further categorized into "homology-driven" (HD) and "non-homology-driven" (nHD) based on normalized homology-based alignment length scores.
  6. Downstream Analyses: The classified SVs were then analyzed for enrichment in various genomic elements (e.g., coding regions, enhancers, TADs), gene sets (e.g., housekeeping genes, disease-related pathways), and their association with population structure and rare disease cohorts.

Datasets

The study utilized several large-scale genomic datasets:

  • Long-Read Whole Genome Sequencing (lrWGS) Data:
    • Human Genome Structural Variation Consortium (HGSVC2) [13]
    • 1000 Genomes Project (1KG) Nanopore data (1KG-ONT) [43]
  • Short-Read Whole Genome Sequencing (srWGS) Data:
    • High-coverage 1KG phase 3 data (1KG) [44]
  • Epigenomic Data: ENCODE project data (ChIP-Seq, DNase-seq, WGB-seq) and 4D Nucleome data (Repli-seq).
  • Rare Disease Cohorts: Data from the 100,000 Genomes (100KG) project for individuals with neurological, ultra-rare, cardiovascular, renal, and endocrine disorders.

The study did not introduce new benchmark datasets but rather leveraged existing comprehensive SV catalogs.

Key Insights and Ideas

  • Beyond Homology: The study powerfully demonstrates that local sequence homology is only one piece of the puzzle in understanding SV origins. Epigenomic context and DNA shape play crucial, often overlooked, roles.
  • DNA Shape and Methylation as Predictors: Features like high local DNA methylation and specific DNA shape characteristics (e.g., low propeller twist variation) are strongly associated with homology-driven SVs.
  • Unsupervised Clustering Uncovers Hidden Patterns: The active-learning unsupervised clustering approach successfully identified SV groups that better reflect underlying repair mechanisms than homology alone, revealing nuanced relationships.
  • Repair Mechanisms and Genomic Context: Different repair pathways (represented by homology-based labels) show distinct preferences for genomic regions (e.g., heterochromatin vs. euchromatin, active vs. quiescent regions).
  • De Novo SVs in Rare Diseases: A notable finding is the distinct upstream H3K27me3 levels observed in de novo SVs from individuals with ultra-rare disorders, suggesting potential epigenetic alterations contributing to disease.
  • Recombination Hotspots and CIHD SVs: Homology-driven SVs (CIHD) are enriched in regions with high recombination rates, particularly in sub-telomeric areas, hinting at a link between recombination machinery and SV formation.

Key Results

  • The homology-based workflow classified SVs into HLH, ILH, and NLH categories, with NLH being the most prevalent, consistent with the high frequency of NHEJ.
  • The unsupervised clustering approach identified distinct SV clusters that correlated with homology-based labels but also revealed SVs with similar repair contexts despite differing homology.
  • Features like DNA methylation and minor groove width (MGW) showed significant differences between homology-driven (HLH) and non-homology-driven (NLH) SVs. For instance, HLH SVs exhibited high normalized mean methylation and low normalized MGW variation.
  • Analysis of rare SVs revealed that CIHD SVs were consistently found in regions with high recombination rates, and these patterns showed population-specific clines, particularly in Europe and Africa.
  • Comparison of de novo SVs from healthy individuals and rare disease cohorts showed distinct upstream H3K27me3 levels in de novo SVs from individuals with ultra-rare disorders.

Limitations and Future Work

  • Reference Genome Bias: The study acknowledges that using reference genome sequences for homology calculations might overlook haplotype-specific variations. Future work could benefit from locally assembled and consensus sequences.
  • SV Type Limitation: The primary focus was on insertions and deletions. While inversions were analyzed, their small sample size and homogeneity prevented detailed comparisons. Future work should broaden the scope to include complex SVs.
  • De Novo SV Rate in 1KG: The authors note that the de novo SV rate in the 1KG dataset might be artificially high due to the use of lymphoblastoid cell lines (LCLs), potentially affecting the accurate representation of germline de novo SVs.
  • Epigenetic Proxy: The use of H1-hESC epigenomic data as a proxy for germline SVs assumes conserved chromatin profiles. Future studies could employ multi-omics data (Hi-C, ATAC-Seq) from affected and unaffected siblings for more direct comparisons.
  • Repair Kinetics: While the study identifies features associated with repair, directly verifying DNA interactions before and after HR and NHEJ to elucidate repair kinetics requires further experimental work.
  • Clinical Phenotype Correlation: While differences were observed in rare disease cohorts, larger sample sizes are needed to bolster clinical distinctions and confirm the role of SVs in disease etiology.

Summary updated

Abstract

The recent advent of long-read whole genome sequencing has enabled us to create an accurate telomere-to-telomere reference genome, construct pangenome graphs, and compile precise catalogs of genomic structural variations (SVs). These comprehensive SV repositories provide an excellent opportunity to explore the role of SVs in genotype-phenotype associations and examine the mechanisms by which SVs are introduced through double-strand break (DSB) repair. Here, we employed comprehensive SV catalogs identified through various short- and long-read whole genome sequencing efforts to infer the underlying mechanisms of SV introduction based on their genomic and epigenomic profiles. Our findings indicate that high local DNA methylation and DNA shape-related features, such as low variations in propeller twist, support the origins of homology-driven SVs. Subsequently, we utilized an active-learning-based unsupervised clustering approach, revealing that homology-dependent SVs show greater evidence of retaining ancestral recombination patterns compared to their homology-independent counterparts. Finally, our comparison of inherited and de novo SVs from healthy populations and rare disease cohorts showed distinct upstream H3K27me3 levels in de novo SVs from individuals with ultra-rare disorders. These findings highlight genome-wide characteristics that may influence the choice of repair mechanisms linked to heritable SV origins.
BibTeX
@inproceedings{boev2025dna,
  author = {Nadejda Boev and Mark Gerstein and Sushant Kumar},
  title = {{DNA shape and epigenomics distinguish the mechanistic origin of human genomic structural variations}},
  booktitle = {Nucleic},
  year = {2025},
  doi = {10.1093/nar/gkaf1325},
}
Figure