Identification of transcription factor co-binding patterns with non-negative matrix factorization.

Area of Science:

• Genomics
• Molecular Biology
• Bioinformatics

Background:

• Transcription factor (TF) binding to DNA is fundamental for regulating gene expression.
• Understanding cooperative TF binding to DNA is crucial but challenging.
• Existing methods lack comprehensive approaches for identifying TF co-binding patterns.

Purpose of the Study:

• To develop and validate a novel computational method, COBIND, for automated identification of TF co-binding patterns.
• To analyze TF co-binding across multiple species and TF families.
• To assess the functional relevance and evolutionary conservation of identified co-binding patterns.

Main Methods:

• COBIND utilizes non-negative matrix factorization (NMF) on DNA regions flanking known TF binding sites (TFBSs).
• The method processes one-hot encoded sequence data to detect enriched DNA patterns at specific distances.
• COBIND was applied to a large dataset (5699 TFBSs) across 401 TFs in seven species.

Main Results:

• COBIND successfully identified known and novel TF co-binding patterns, supported by motif similarity and protein-protein interaction data.
• Across species, 67% of TFs shared co-binding motifs within the same structural families.
• Predicted co-binding patterns exhibited higher evolutionary conservation and were validated by open chromatin and single-molecule footprinting data.

Conclusions:

• COBIND is an effective tool for discovering TF co-binding configurations.
• Co-binding motifs are prevalent among structurally related TFs, suggesting conserved regulatory mechanisms.
• The identified co-binding patterns are functionally significant and evolutionarily conserved, providing insights into transcription regulation.