GPS

GPS identifies and precisely localizes protein–DNA interaction events in ChIP-seq datasets, resolving closely spaced homotypic binding events in invertebrate and mammalian promoters and enhancers.

Key Features:

• High-resolution peak detection: Uses a high spatial resolution peak detection algorithm to predict precise protein–DNA interaction sites from ChIP-seq reads.
• Complexity-penalized mixture model: Models observed reads with a complexity-penalized mixture model to balance model fit and complexity.
• Segmented Expectation-Maximization (EM): Employs an efficient segmented EM algorithm to estimate event locations.
• Homotypic event resolution: Accurately resolves closely spaced homotypic events that may appear as single read clusters due to proximity.
• Accounts for ChIP fragmentation variability: Addresses random variations from the ChIP fragmentation process that can obscure event locations.
• Cross-dataset event alignment (optional): Provides an optional mode to align common events across distinct experimental datasets.
• Empirical performance: Demonstrated superior detection of joint events and enhanced spatial resolution, with comparable or superior specificity and sensitivity in comparative analyses on synthetic and real ChIP-seq data.

Scientific Applications:

• Promoter and enhancer analysis: Mapping protein–DNA interactions in invertebrate and mammalian promoters and enhancers.
• Detection of joint events: Resolving and quantifying closely spaced, homotypic transcription factor binding events.
• Regulatory mechanism studies: Enabling precise localization of interactions to support investigations of gene regulation.

Methodology:

Observed reads are modeled using a complexity-penalized mixture model, with event locations estimated by a segmented Expectation-Maximization algorithm and an optional mode for aligning common events across datasets; peak detection uses a high spatial resolution algorithm.