Uncovering the Relationship between Tissue-Specific TF-DNA Binding and Chromatin Features through a Transformer-Based...

Area of Science:

• Bioinformatics
• Genomics
• Molecular Biology

Background:

• Chromatin features offer insights into tissue-specific transcription factor-DNA binding, crucial for physiological processes.
• Identifying transcription factor-DNA binding and its link to chromatin features is a key bioinformatics challenge due to complex mechanisms and data heterogeneity.

Purpose of the Study:

• To develop a robust model, GHTNet (General Hybrid Transformer Network), for predicting transcription factor-DNA binding specificity.
• To decode the relationship between tissue-specific transcription factor-DNA binding and chromatin features.
• To interpret the interplay of transcription factors, chromatin features, and diseases.

Main Methods:

• Developed GHTNet, a transformer-based network utilizing an alternative input scheme.
• Applied GHTNet to analyze TF-DNA binding specificity and its relationship with chromatin features.
• Conducted cross-species studies to address data limitations.

Main Results:

• GHTNet achieved a 5% absolute performance improvement over existing methods in predicting TF-DNA binding specificity.
• Analysis revealed that DNA sequence is the most important predictor, followed by epigenomics and shape, with tissue-specific variations.
• Demonstrated GHTNet's utility in interpreting relationships among TFs, chromatin features, and AD46 tissue-associated diseases.

Conclusions:

• GHTNet is an accurate and robust framework for deciphering tissue-specific TF-DNA binding.
• The model provides insights into gene regulation, non-coding region interpretation, and disease associations.
• Cross-species analysis offers novel strategies for handling limited data in TF-DNA binding studies.