Allele-Specific Folding

The foundational step in allele-specific folding analysis, requiring the assignment of sequencing reads from chromosome conformation capture assays to either the maternal or paternal allele based on heterozygous single nucleotide polymorphisms (SNPs). This process generates haplotype-resolved contact maps that reveal divergent folding patterns between homologous chromosomes. Phasing algorithms integrate read-level variant information with proximity ligation data, often employing statistical models like HapCUT2 or WhatsHap to resolve haplotypes across entire chromosomes before downstream structural comparison.

Heterozygous genetic variants can create asymmetric loop extrusion dynamics between maternal and paternal alleles. Key mechanisms include:

• CTCF motif disruption: A single nucleotide variant in a CTCF binding site on one allele can abolish loop anchoring, causing the loop to form exclusively on the intact allele
• Cohesin loading bias: Structural variants near cohesin loading sites alter the frequency or directionality of extrusion on one haplotype
• Differential boundary strength: Insertions or deletions that modify the affinity of TAD boundary elements lead to allele-specific insulation patterns These imbalances are quantified using allelic contact ratios comparing interaction frequencies at loop anchors between haplotypes.

Large-scale genomic rearrangements, such as copy number variations or inversions, can cause one allele to shift between the active A compartment and the inactive B compartment. This switching is detected by comparing the long-range interaction eigenvector profiles of each haplotype. For example, a deletion removing an enhancer cluster on the maternal allele may cause surrounding regions to transition from A to B, while the intact paternal allele maintains active compartment status. Compartment strength scores and saddle plot asymmetry metrics quantify the degree of allelic divergence in chromatin state.

Topologically Associating Domain boundaries can be haplotype-specific when boundary-defining elements harbor heterozygous variants. A single base change in a CTCF motif can weaken or eliminate boundary function on one allele, causing TAD merging and aberrant enhancer-promoter contacts. Computational detection involves:

• Comparing insulation score profiles between haplotypes at each genomic bin
• Identifying differential boundary strength using permutation-based significance testing
• Validating with allele-aware directionality index calculations These allele-specific boundary disruptions are a primary mechanism by which non-coding variants in genome-wide association study loci exert pathogenic effects.

Genomic imprinting creates a natural form of allele-specific folding where epigenetic marks on one parental allele silence genes and alter local chromatin architecture. Imprinted loci exhibit distinct 3D structures between maternal and paternal chromosomes, often involving allele-specific insulator function mediated by differentially methylated regions. Deep learning models trained to predict folding must account for these parent-of-origin effects by incorporating DNA methylation status and histone modification data as allele-specific input features, rather than relying solely on DNA sequence.

Modern sequence-to-structure models like Akita and allele-aware extensions of DeepHiC predict haplotype-resolved contact maps by processing maternal and paternal DNA sequences through separate forward passes. Key architectural considerations include:

• Siamese network designs that share weights between allele-specific branches to learn conserved folding principles while capturing variant-driven differences
• Attention mechanisms that focus on heterozygous positions to model their disproportionate structural impact
• Contrastive loss functions that explicitly optimize the model to distinguish allele-specific contact patterns These models enable in silico prediction of how rare variants alter 3D genome organization without requiring experimental Hi-C data for every genotype.

The computational process of assigning genetic variants to their parental chromosome of origin. Statistical phasing uses population reference panels, while trio-based phasing leverages parental genotypes. Phasing is a prerequisite for allele-specific folding analysis, as it determines which variants co-occur on the same homolog. Without accurate phasing, distinguishing maternal from paternal chromatin interaction patterns is impossible.

A genome-wide matrix quantifying interaction frequencies between all pairs of genomic loci, derived from chromosome conformation capture assays. In allele-specific contexts, Hi-C reads overlapping heterozygous variants are assigned to maternal or paternal haplotypes, producing haplotype-resolved contact maps. These maps reveal how structural variants and single nucleotide polymorphisms differentially shape the 3D architecture of each homolog.

A self-interacting genomic region where DNA sequences physically interact more frequently with each other than with sequences outside the domain. Allele-specific TAD analysis examines whether heterozygous structural variants alter domain boundaries on one homolog. Deletions spanning boundary elements can cause TAD fusion, while boundary-creating insertions can split a single TAD into two, leading to allelic imbalance in gene regulation.

The computational task of identifying physical contacts between distal enhancer elements and gene promoters. In allele-specific folding, models predict how heterozygous variants in enhancers or boundary elements alter these interactions. A single nucleotide variant in a CTCF motif can disrupt loop anchoring, causing an enhancer to skip its target promoter and activate a neighboring gene instead—a mechanism implicated in allele-specific disease susceptibility.

The computational assessment of how large-scale genomic rearrangements alter 3D genome folding. Allele-specific models evaluate how deletions, inversions, or duplications on one homolog disrupt normal chromatin architecture. Key considerations include:

• TAD boundary disruption leading to ectopic enhancer-promoter contacts
• Loop domain reorganization creating novel regulatory landscapes
• Copy number variation altering interaction stoichiometry These predictions are critical for interpreting non-coding variants in rare disease diagnostics.