Predicting transcription factor affinities to DNA from a biophysical model

Area of Science:

• Molecular Biology
• Bioinformatics
• Genomics

Background:

• Traditional gene expression studies focus on transcription factor binding sites.
• Recent advances incorporate experimental data on protein binding affinities to intergenic sequences.
• This work investigates the convergence of theoretical and experimental approaches to gene regulation.

Purpose of the Study:

• To develop and validate a physical binding model for predicting transcription factor (TF) relative binding affinities.
• To assess the model's performance against existing methods for genome-wide binding data analysis.
• To identify biologically meaningful parameters for predicting TF binding in the absence of experimental data.

Main Methods:

• Utilized a physical binding model incorporating simple count matrices.
• Employed a two-parameter model to predict relative binding affinities.
• Applied the model to genome-wide binding data in yeast.

Main Results:

• A significant portion of yeast genome-wide binding data was explained by the model.
• The physical model proved more powerful than traditional methods requiring cutoff selection.
• Biologically meaningful parameters were derived, enabling prediction of binding affinities.

Conclusions:

• The developed physical model effectively predicts transcription factor binding affinities.
• This approach enhances the understanding of gene expression regulation.
• The TRAP program is available for non-commercial use.