MATCHA

MATCHA analyzes multi-way chromatin interaction data using hypergraph representation learning to represent complex multi-locus contacts and support studies of higher-order genome organization and gene regulation at single-nucleus resolution.

Key Features:

• Hypergraph Representation Learning: Represents multi-way chromatin interactions as hyperedges in a hypergraph to model complex multi-locus contact patterns.
• Data Denoising and Prediction: Applies representation learning to denoise noisy SPRITE and ChIA-Drop datasets and to make de novo predictions of multi-way interactions.
• Application to Ligation-free Genome-wide Mapping: Processes ligation-free, genome-wide chromatin interaction data from SPRITE and ChIA-Drop to handle high-resolution multi-way contacts.

Scientific Applications:

• Higher-order Chromosome Organization: Characterizes higher-order chromosome organization by capturing multi-way contacts among multiple genomic loci.
• Gene Regulation Mechanisms: Supports investigation of gene regulation mechanisms by resolving multi-locus interactions at single-nucleus resolution.
• Refinement of Interaction Maps: Refines and augments multi-way interaction maps derived from SPRITE and ChIA-Drop through denoising and prediction.

Methodology:

Transforms multi-way interaction data into a hypergraph where each hyperedge represents a multi-locus interaction and applies representation learning to denoise data and predict new interactions.