Precise physical models of protein-DNA interaction from high-throughput data

Area of Science:

• Molecular Biology
• Systems Biology
• Computational Biology

Background:

• Gene transcription regulation relies on transcription factor (TF) binding to DNA.
• Accurate modeling of TF-DNA binding energy is crucial for quantitative biology.
• High-throughput binding assays generate data for TF-DNA interaction studies.

Purpose of the Study:

• To develop a principled, likelihood-based approach for inferring physical models of TF-DNA binding energy.
• To enable probabilistic predictions of TF binding sites and their binding energies.
• To validate the inferred models using existing experimental and phylogenetic data.

Main Methods:

• A likelihood-based computational framework for TF-DNA binding energy modeling.
• Computation of likelihood without explicit noise assumptions in binding assay data.
• Probabilistic sampling of model parameters to predict binding sites and energies.
• Application to Saccharomyces cerevisiae TF Abf1p data.

Main Results:

• Inferred TF binding models with precisely determined parameters.
• High degree of phylogenetic conservation in predicted binding site energies.
• Consistent characterizations of Abf1p from in vivo and in vitro experiments.
• Validation of the model's accuracy through independent data.

Conclusions:

• The developed likelihood-based method accurately models TF-DNA binding energy.
• The approach provides precise predictions of binding sites and their physical energies.
• Phylogenetic and experimental data strongly support the accuracy and consistency of the models.
• This method offers a robust tool for quantitative analysis of TF-DNA interactions.