Modeling associated protein-DNA pattern discovery with unified scores

Area of Science:

• Genomics
• Bioinformatics
• Molecular Biology

Background:

• Understanding protein-DNA interactions, particularly transcription factor (TF) and transcription factor binding site (TFBS) binding, is fundamental to deciphering gene regulation.
• Current methods for TF-TFBS pattern discovery often focus on one side (TF or TFBS) or rely on existing annotations, lacking a unified approach to prioritize verifiable patterns.

Purpose of the Study:

• To propose a formal, unified (both-sided) model for associated TF-TFBS pattern discovery.
• To develop an effective pipeline for identifying and prioritizing top verifiable TF-TFBS patterns using sequence data.

Main Methods:

• Developed a novel pipeline for associated TF-TFBS pattern discovery using sequence data only.
• Introduced unified scoring to prioritize patterns based on both TF and TFBS information.
• Conducted instance-level evaluations and literature surveys for verification.

Main Results:

• The proposed unified scoring method significantly improves the accuracy of identifying protein-DNA binding cores compared to previous works, with up to 90% verification for top-ranked patterns.
• Extended verification through literature surveys showed a high correlation between unified scores and confirmed TF-TFBS interactions.
• Top scored patterns successfully matched known WRKY binding cores and aligned with high binding affinities from in vivo experiments, even without 3D structural data.

Conclusions:

• The unified model and pipeline provide a robust method for discovering and prioritizing associated TF-TFBS patterns.
• This approach enhances the understanding of gene regulation and offers promising applications in subtype and disease analysis.
• The findings are validated through structural data, literature, and in vivo experimental results, demonstrating broad applicability.