AREM

AREM applies an expectation-maximization probabilistic framework to map short ChIP-Seq reads — including multi-mapped reads — to a reference genome and identify genome-wide enriched regions (peaks) for characterization of protein–DNA interactions.

Key Features:

• Probabilistic Approach: Employs probabilistic models that incorporate all sequencing reads, including reads that map to multiple genomic locations.
• Expectation-Maximization Algorithm: Implements an iterative Expectation-Maximization (E-M) algorithm that updates alignment probabilities for reads across genomic locations.
• Mixture Model Framework: Models reads with a mixture model that distinguishes K enriched regions from a null genomic background.
• Enhanced Detection Power: Increases power to detect binding events, particularly within repeat sequences, by leveraging multi-mapped reads rather than restricting to uniquely mapped reads.

Scientific Applications:

• Genome-wide protein–DNA interaction characterization: Enables genome-wide characterization of protein–DNA interactions for transcription factors, cofactors, chromatin modifiers, and other DNA-binding proteins.
• Rad21 peak identification (mouse): Applied to the mouse genome to identify 19,935 Rad21 peaks, including 7.6% of peaks missed by methods using only uniquely mapped reads.
• Srebp-1 peak identification (mouse): Applied to the mouse genome to detect 1,748 Srebp-1 peaks, of which 13% were identified only when multi-mapped reads were considered.

Methodology:

Aligns short ChIP-Seq reads to a reference genome using probabilistic models, models reads with a mixture model of K enriched regions versus a null background, and applies an iterative Expectation-Maximization algorithm to update alignment probabilities and identify enriched regions (peaks).