GRiNCH

GRiNCH applies graph-regularized non-negative matrix factorization and clustering to smooth sparse chromatin contact count matrices from Hi-C and detect topologically associating domains (TADs).

Key Features:

• Graph-regularized NMF: Integrates graph regularization with non-negative matrix factorization (NMF) to model chromatin contact structure.
• Constrained matrix factorization: Performs constrained matrix factorization that simultaneously smooths sparse contact count matrices and discovers TADs.
• Clustering: Uses clustering techniques to define domain boundaries and group interacting genomic regions.
• Sparse data handling: Operates on sparse chromatin contact count matrices typical of Hi-C assays.
• Cross-platform applicability: Applicable to chromatin interaction data from SPRITE and HiChIP in addition to Hi-C.
• Boundary factor prediction: Predicts novel boundary factors that may be associated with genome organization.
• Comparative performance: Demonstrated superior performance relative to seven other TAD-calling algorithms and three smoothing techniques in comparative analyses.

Scientific Applications:

• TAD detection: Detects topologically associating domains (TADs) from Hi-C contact matrices.
• Chromatin architecture analysis: Analyzes genome organization using Hi-C, SPRITE, and HiChIP chromatin interaction data.
• Boundary factor discovery: Identifies candidate boundary factors to inform studies of genome organization and gene regulatory mechanisms.

Methodology:

Graph-regularized non-negative matrix factorization (NMF), constrained matrix factorization for simultaneous smoothing and domain discovery, and clustering techniques.