CLNN-loop

CLNN-loop predicts chromatin loops mediated by the CCCTC-binding factor (CTCF) from diverse sequence-based features to enable analysis of 3D genome organization and CTCF-binding site (CBS) interactions.

Key Features:

• Deep Learning Architecture: Employs an advanced deep learning framework to model sequence determinants of CTCF-mediated chromatin loops.
• Sequence-based Feature Integration: Integrates multiple sequence-derived features to inform loop prediction for different CBS pair types.
• Cross-cell-line and CBS-pair Generalization: Demonstrates predictive performance across various cell lines and types of CTCF-binding site (CBS) pairs.
• Interpretability via SHAP: Uses SHAP (SHapley Additive exPlanations) to quantify feature contributions, highlighting CTCF motifs and sequence conservation as key indicators.
• Performance Superiority: Shows higher accuracy and generalization ability compared with existing methods for chromatin loop prediction.

Scientific Applications:

• 3D Genome Organization: Enables computational mapping of chromatin loop architecture mediated by CTCF across cell types.
• Gene Regulation Studies: Supports investigation of regulatory interactions linked to chromatin looping and CTCF binding.
• Epigenetics Research: Facilitates analysis of how sequence features associated with CBS contribute to epigenetic regulation via looping.
• Disease Mechanism Exploration: Aids study of molecular underpinnings of diseases associated with altered genomic architecture.
• Genomics and Personalized Medicine: Provides predictive insights that can inform genomics research relevant to personalized medicine.

Methodology:

The model integrates diverse sequence-based features and applies deep learning to predict CTCF-mediated chromatin loops, focusing on interactions between different CBS pair types across cell lines, and uses SHAP analysis to interpret feature contributions.