TF-loop: deciphering the transcription factor regulatory language for CTCF-mediated chromatin loop based on BERT.

Area of Science:

• Genomics
• Computational Biology
• Molecular Biology

Background:

• Chromatin looping is crucial for 3D genome organization and gene expression regulation.
• Transcription factors (TFs), especially CTCF and Cohesin, dynamically regulate chromatin loop formation.
• Current computational methods for predicting CTCF-mediated loops face challenges with genome-wide predictions and imbalanced datasets.

Purpose of the Study:

• To develop a novel computational framework for accurate prediction of chromatin loops.
• To address limitations of existing DNA-sequence-based models in capturing TF binding dynamics.
• To leverage natural language processing for decoding the regulatory code of chromatin looping.

Main Methods:

• Introduced TF-loop, a TF regulatory language framework.
• Conceptualized TF sequences as a structured language based on TF binding positions and orientations.
• Utilized the BERT model to decode linguistic patterns within TF sequences for loop prediction.

Main Results:

• TF-loop significantly improves the accuracy of chromatin loop prediction compared to state-of-the-art models.
• The framework demonstrates robust performance across diverse cell types, even with imbalanced training data.
• Successfully decoded latent linguistic patterns in TF sequences to predict chromatin loops.

Conclusions:

• TF-loop offers a novel perspective on decoding 3D genome structure using natural language processing.
• The framework enhances the prediction of TF-mediated chromatin loops, aiding gene expression studies.
• Highlights the potential of TF-loop for advancing genomic research and understanding gene regulation.