FitHiC

FitHiC assigns statistical confidence estimates to chromosomal contact maps from Hi-C and related genome architecture assays to identify significant chromatin contacts.

Key Features:

• Statistical Confidence Estimation: Computes statistical confidence estimates for Hi-C contact maps using a monotonically non-increasing spline to model the relationship between genomic distance and contact probability without relying on parametric assumptions.
• Bias Correction: Corrects contact probabilities by accounting for bin- or locus-specific biases and covariates that impact Hi-C contact counts.
• Scalability and Resolution Flexibility: Supports genome-wide analysis of intra-chromosomal distances and inter-chromosomal contacts across resolutions including mid-range (20 kb–2 Mb) and single restriction cut site resolution.
• Merging Filter Module: Applies a merging filter to eliminate indirect or bystander interactions, reducing reported contacts while preserving key chromatin loops such as those between convergent CTCF binding sites.
• Use Cases and Performance: Demonstrated on 5-kb GM12878, 40-kb IMR90, and single restriction cut site budding yeast Hi-C datasets with reported sequential and parallel runtimes for preprocessing and analysis.
• Resource Efficiency: Described resource requirements for high-performance computing environments, including typical processor and memory footprints and peak memory for 1 kb Hi-C analysis.

Scientific Applications:

• Chromatin architecture analysis: Identification of significant chromatin contacts and loops to study three-dimensional genome organization.
• Regulatory interaction mapping: Detection of interactions between regulatory elements to inform studies of gene regulation.
• Loop characterization: Preservation and identification of specific chromatin loops, including those anchored by convergent CTCF sites.
• Cross-resolution and species analyses: Analysis of Hi-C data across resolutions and organisms, including human cell lines (GM12878, IMR90) and budding yeast datasets.

Methodology:

Uses a monotonically non-increasing spline to model distance-dependent contact probability, performs bias correction for bin- or locus-specific covariates, and applies a merging filter module to remove indirect/bystander interactions while supporting intra- and inter-chromosomal contact analysis across variable resolutions.