ShNeigh

ShNeigh reconstructs chromatin 3D structures from Hi-C contact frequency matrices by incorporating adjacency relationships between neighboring genomic loci to improve inference of spatial genome organization.

Key Features:

• Adjacency relationship consideration: Incorporates the adjacency relationship between neighboring genomic loci, modeling their mutual dependencies in the polymer structure.
• Hi-C input: Uses genome-scale contact frequency matrices from high-throughput 3C technologies such as Hi-C as input.
• MDS combined with local dependence model: Combines classical Multidimensional Scaling (MDS) with a local dependence model for neighboring loci represented by a Gaussian formula.
• Genomic-distance-weighted smoothness constraint: Imposes a smoothness constraint in the MDS optimization weighted by a function of genomic distance, penalizing large deviations inconsistent with adjacency.
• Inference from noisy and incomplete data: Infers accurate 3D chromatin structures from noisy and incomplete Hi-C contact frequency matrices.
• Efficiency: Maintains the high-speed computational properties of classical MDS.
• Accuracy and robustness: Demonstrates superior accuracy and robustness compared to existing methods, particularly with sparse contact maps.
• Stability in high-noise settings: Recovers stable chromatin structures under high-noise conditions.
• Versatility across restriction enzymes: Reconstructs similar 3D structures from Hi-C datasets generated using different restriction enzymes.

Scientific Applications:

• Spatial organization of chromosomes: Enables reconstruction of chromatin 3D structures to study the spatial organization of chromosomes.
• Gene regulation and disease mechanisms: Supports investigation of how genomic architecture influences gene expression, biological processes, and disease mechanisms.

Methodology:

Operates on Hi-C contact frequency matrices using classical Multidimensional Scaling (MDS) integrated with a local Gaussian dependence model for neighboring loci and a genomic-distance-weighted smoothness constraint in the MDS optimization.