Peakachu

Peakachu predicts chromatin loops from genome-wide interaction matrices such as Hi-C using supervised machine learning.

Key Features:

• Supervised learning integration: Integrates orthogonal interaction datasets including ChIA-PET, HiChIP, Capture Hi-C, and high-throughput imaging to inform loop prediction.
• Random Forest classification: Employs Random Forest classifiers to distinguish true chromatin interactions from background signal.
• Cross-platform performance: Demonstrates predictive consistency across multiple platforms, sequencing depths, and species.
• Short-range interaction detection: Exhibits improved sensitivity for identifying short-range chromatin interactions relative to enrichment-based approaches.

Scientific Applications:

• Gene regulation: Mapping chromatin loops to study regulatory contacts between promoters, enhancers, and other regulatory elements.
• Genome organization: Characterizing three-dimensional genome architecture and spatial configuration of the genome.
• Disease and cellular function studies: Investigating how chromatin interactions relate to cellular function and disease mechanisms.

Methodology:

Random Forest classifiers are trained on labeled datasets derived from interaction matrices and orthogonal data types (ChIA-PET, HiChIP, Capture Hi-C, high-throughput imaging) and then applied to predict chromatin loops in new datasets, including application to 56 Hi-C datasets.