ChromDMM

ChromDMM applies a product Dirichlet-multinomial mixture model to cluster genomic regions characterized by multiple chromatin features, enabling analysis of heterogeneous epigenetic data at regulatory elements.

Key Features:

• Product Dirichlet-multinomial mixture model: Models multivariate count data from multiple chromatin features using a product of Dirichlet-multinomial components to capture overdispersion and feature-specific signal.
• Profile Shifting and Flipping: Probabilistically accounts for positional inaccuracies and strand-orientation by shifting and flipping profiles during clustering.
• Hyper-parameter Optimization: Regularizes the smoothness of epigenetic profiles across consecutive genomic regions via hyper-parameter tuning.
• Enhanced Clustering Accuracy: Demonstrates superior performance in clustering, shifting, and strand-orienting profiles in simulation-based comparisons to previous methods.

Scientific Applications:

• Epigenetic and chromatin feature analysis: Clusters and characterizes epigenetic modifications and chromatin features at genomic regulatory elements to reveal distinct signal patterns.
• Enhancer region characterization: Identifies distinct chromatin feature patterns at enhancer regions that can be validated by enrichment for transcriptional regulatory factor binding sites.

Methodology:

Implements a product Dirichlet-multinomial mixture model with probabilistic profile shifting and flipping to handle positional and strand-orientation inaccuracies, employs hyper-parameter optimization to enforce profile smoothness across regions, and uses simulations for performance evaluation.