msCentipede

msCentipede extends the CENTIPEDE probabilistic framework to infer transcription factor binding sites by modeling DNase-seq and paired-end ATAC-seq chromatin accessibility cleavage profiles together with genomic sequence content.

Key Features:

• Probabilistic Framework: msCentipede employs a probabilistic model that combines chromatin hypersensitivity signatures, spatial DNase/ATAC cleavage profiles, and genomic sequence content to infer transcription factor binding.
• Multi-Scale Models: Implements multi-scale models of inhomogeneous Poisson processes to account for variation in cleavage patterns within and across binding sites.
• Spatial Structure Analysis: Captures spatial heterogeneity in DNase I and ATAC-seq cleavage patterns specific to each transcription factor to improve binding-site localization.
• Replicate Data Utilization: Incorporates variation observed across replicate DNase I experiments to model between-replicate heterogeneity.
• Sequence Bias Adjustment: Includes a background model extension to mitigate DNase I sequence bias effects on footprinting-based inference.
• Versatility with Data Types: Applicable to chromatin accessibility datasets including DNase-seq and paired-end ATAC-seq.

Scientific Applications:

• Gene Regulation Studies: Annotates regulatory elements and cell-type–specific transcription factor binding to support analyses of gene regulation.
• Comparative Analysis: Enables comparison of inferred binding sites to Chip-seq peaks to validate and refine transcription factor activity models.
• Research on Lymphoblastoid Cell Lines: Demonstrated improved TFBS inference using DNase-seq measurements in lymphoblastoid cell lines.

Methodology:

Extends the CENTIPEDE probabilistic framework using multi-scale modeling of inhomogeneous Poisson processes, refines spatial cleavage-profile models, incorporates replicate variation, includes a background model for DNase I sequence bias, and is implemented in Python.