MaxHiC

MaxHiC applies statistical modeling to Hi-C and capture Hi-C data to correct experimental biases and identify statistically significant chromatin interactions for studying gene regulation and disease-associated regulatory architectures.

Key Features:

• Background Correction: Corrects for biases including genomic distance, GC content, and mappability to improve accuracy of interaction frequency estimates.
• Statistical Significance: Models interaction counts using a negative binomial distribution with parameters estimated by maximum likelihood to assign statistical significance to interactions.
• Applicability Across Data Types: Implements distinct models optimized for both Hi-C and capture Hi-C datasets.
• Performance and Benchmarking: Benchmarking against other Hi-C background correction tools reported higher overlap of MaxHiC-identified interactions with active chromatin histone marks, CTCF binding sites, DNase sensitivity regions, and disease-associated GWAS SNPs.
• Biological Relevance: Links interacting regions to expression quantitative trait loci (eQTL) pairs and known regulatory features to support interpretation of regulatory mechanisms.
• Scalability: Scales to analyze very deep Hi-C libraries, including high-resolution MicroC datasets.

Scientific Applications:

• Chromatin architecture analysis: Identification of higher-order chromatin structures and significant chromatin interactions from Hi-C and capture Hi-C data.
• Regulatory element annotation: Annotation and prioritization of interactions overlapping active histone marks, CTCF binding sites, and DNase hypersensitive regions.
• eQTL and gene regulation studies: Linking distal regulatory elements to target genes through overlap with eQTL pairs and interaction evidence.
• Disease genetics: Prioritizing disease-associated regulatory interactions by integrating GWAS SNPs with significant chromatin contacts.

Methodology:

Uses a negative binomial distribution with maximum likelihood estimation, applies simultaneous bias correction for genomic distance, GC content, and mappability, employs distinct models for Hi-C and capture Hi-C datasets, and leverages statistical modeling and machine learning techniques with benchmarking against other background-correction methods.