Inferring CTCF-binding patterns and anchored loops across human tissues and cell types.

Area of Science:

• Genomics
• Epigenetics
• Computational Biology

Background:

• CCCTC-binding factor (CTCF) is a key transcription regulator with poorly understood roles in DNA sequence recognition, chromosome barrier function, and enhancer blocking.
• Current computational tools are limited in assessing CTCF-binding sites' regulatory potential and their impact on chromatin loop formation.

Purpose of the Study:

• To develop a deep-learning model for accurate characterization of CTCF binding using high-resolution genomic and epigenomic features.
• To identify distinct chromatin and sequence patterns associated with CTCF-mediated insulation and looping.
• To establish a compendium of CTCF-anchored loops and investigate their role in disease.

Main Methods:

• Development of a deep-learning model, DeepAnchor, for CTCF binding characterization.
• Utilizing high-resolution genomic and epigenomic data as input features.
• Optimization of a loop prediction model using DeepAnchor scores.
• Creation of a compendium of CTCF-anchored loops across 52 human tissue/cell types.

Main Results:

• DeepAnchor accurately characterizes CTCF binding, revealing distinct sequence and chromatin patterns for insulation and looping.
• An optimized loop model based on DeepAnchor scores demonstrates high accuracy in predicting CTCF-anchored loops.
• A comprehensive compendium of CTCF-anchored loops across diverse human cell types was established.

Conclusions:

• DeepAnchor provides a powerful tool for understanding CTCF-mediated gene regulation.
• Genomic disruption of CTCF-anchored loops may represent a common mechanism in disease pathogenesis.
• The developed models and resources facilitate research into CTCF's role in cellular development and disease.