Variable structure motifs for transcription factor binding sites.

Area of Science:

• Genomics
• Bioinformatics
• Molecular Biology

Background:

• Traditional DNA-transcription factor binding site (TFBS) models use fixed-length Position Weight Matrices (PWMs).
• Some transcription factors exhibit flexible binding configurations affecting TFBS function.
• Increasing ChIP-seq data necessitates models that capture this binding flexibility.

Purpose of the Study:

• To develop and evaluate a novel variable-length gapped PWM model for DNA binding sites.
• To investigate transcription factor binding flexibility using ChIP-seq data.
• To provide more interpretable motif models for structural studies.

Main Methods:

• Re-analysis of TRANSFAC database binding sites.
• Application of a novel motif search algorithm to ChIP-seq data.
• Comparison with standard PWM finders and alternative variable structure motif methods.
• Held-out cross-validation tests.

Main Results:

• The new variable-length model showed a better fit for certain binding sites.
• The method recovered known variable structure motifs for p53, Stat5a, and Stat5b.
• A novel, generalized variable-length PWM for Sp1 was discovered, improving classification performance.
• Performance was comparable to existing methods in cross-validation.

Conclusions:

• A new, non-restrictive, and appropriately parameterized gapped PWM model for variable-length DNA binding sites was developed.
• While not universally superior in predictive accuracy, the model yields more interpretable motifs.
• This represents the first application of variable-length motif models to eukaryotic ChIP-seq data, demonstrating their utility.
• A novel motif for the transcription factor Sp1 was identified.