CSREP

CSREP estimates representative chromatin-state probability maps from groups of samples and detects high-resolution differential chromatin-state assignments between sample groups to summarize chromatin state annotations and support epigenetic analyses.

Key Features:

• Probabilistic Estimation: CSREP uses probabilistic models to estimate chromatin-state probabilities at each genomic position and generate representative maps that encapsulate group information.
• Multi-Class Logistic Regression Classifiers: An ensemble of multi-class logistic regression classifiers predicts individual-sample chromatin state assignments using state maps derived from all other samples within the group.
• High-Resolution Differential Analysis: CSREP identifies genomic locations with differential chromatin state assignments between groups by comparing chromatin-state probability assignments across sample sets.

Scientific Applications:

• Non-coding Genome Annotation: Summarizing chromatin state annotations across samples to aid annotation of the non-coding genome.
• Epigenetic Modification Analysis: Detecting differential chromatin states to study epigenetic modifications across cell and tissue types.
• Regulatory Mechanism Investigation: Pinpointing differential chromatin-state assignments to investigate regulatory mechanisms underlying gene expression and cellular differentiation.

Methodology:

CSREP takes input chromatin state annotations from a group of samples, applies probabilistic estimation to generate representative chromatin-state probability maps at each genomic position, employs an ensemble of multi-class logistic regression classifiers that predict individual sample states based on state maps derived from the other samples in the group, and compares chromatin-state probability assignments across groups to identify differential genomic locations.