normr

normr models ChIP-seq read counts to normalize signal and detect significantly enriched or depleted chromatin-associated protein binding sites and histone modifications.

Key Features:

• Binomial mixture model: Models read counts using a binomial mixture model with user-defined components.
• User-defined components: Allows specification of mixture components to represent background and enrichment states.
• Background estimation: Fits a background estimate that accounts for enrichment effects in specific genomic regions to define an appropriate null hypothesis.
• Control-based normalization: Normalizes ChIP-seq read densities by comparing them against control read densities.
• Bias minimization: Reduces technical variation to refine statistical analysis of read data.
• Enrichment calling across SNRs: Handles datasets with diverse signal-to-noise ratios, exemplified by H3K4me3 (high signal) and H3K36me3 (low signal).
• Broad and peak pattern detection: Captures both broad and peak enrichment patterns relevant for histone marks such as H3K27me3 and H3K9me3.

Scientific Applications:

• Enrichment calling: Identification of enriched regions for histone modifications and chromatin-associated proteins across varying signal intensities (e.g., H3K4me3, H3K36me3).
• Chromatin state discovery: Detection of novel heterochromatic regimes characterized by broad and peak enrichment patterns for histone modifications such as H3K27me3 and H3K9me3.
• Differential enrichment analysis: Comparison of enrichment between conditions or cell types, for example HepG2 hepatocarcinoma cells versus primary human hepatocytes for H3K4me3 and H3K27me3.

Methodology:

Read counts are modeled with a binomial mixture model (user-defined components) and ChIP read densities are compared to control read densities to fit a background used as the null hypothesis for calling enriched or depleted regions.