Predicting transcription factor specificity with all-atom models.

Area of Science:

• Computational biology
• Structural biology
• Genetics

Background:

• Transcription factors (TFs) regulate gene expression by binding to specific DNA operator sites.
• Predicting TF binding sites typically requires prior knowledge of known binding sequences.
• An ab initio approach, independent of prior binding data, is needed for broader TF binding site prediction.

Purpose of the Study:

• To investigate the feasibility of using structure-based energy calculations for ab initio prediction of TF DNA binding sites.
• To evaluate the accuracy of atomistic models in distinguishing cognate from noncognate DNA sites for the PurR transcription factor from Escherichia coli.
• To compare the performance of this structure-based approach against traditional bioinformatic methods.

Main Methods:

• Utilized structure-based energy calculations on protein-DNA complexes without prior knowledge of bound sites.
• Studied the PurR transcription factor from Escherichia coli.
• Systematically evaluated the approach by comparing it with bioinformatic methods and testing against random DNA sequences and homologous TFs.
• Analyzed the impact of experimental mutations in both DNA and the protein.

Main Results:

• Structure-based energy calculations can effectively distinguish between cognate and noncognate DNA sites for the PurR TF.
• The specificity of PurR binding is primarily determined by direct protein-DNA interactions.
• DNA bending plays a minimal role in the specificity of PurR binding.

Conclusions:

• Structure-based energy calculations offer a viable ab initio approach for predicting transcription factor binding sites.
• Direct protein-DNA interactions are the dominant factor in determining TF binding specificity.
• This method advances computational prediction of gene regulation mechanisms.