Combining Physics-Based Protein-DNA Energetics with Machine Learning to Predict Interpretable Transcription...

Area of Science:

• Computational biology
• Biophysics
• Genomics

Background:

• Transcription factors (TFs) regulate gene expression by binding to specific DNA sequences.
• Alterations in TF-DNA binding affinity and specificity are linked to diseases like cancer and developmental disorders.
• Accurate prediction of TF-DNA interactions is crucial for understanding gene regulation and disease.

Purpose of the Study:

• To develop a computational framework integrating physics-based simulations and machine learning (ML) for predicting protein-DNA binding affinities and specificities.
• To enhance the accuracy and interpretability of TF-DNA binding predictions.
• To investigate the key molecular determinants of TF-DNA binding affinity and specificity.

Main Methods:

• Combined all-atom molecular dynamics (MD) simulations and Molecular Mechanics-Generalized Born Surface Area (MMGBSA) calculations.
• Employed machine learning models (neural networks, random forests, support vector machines) for prediction.
• Utilized high-quality experimental data from genomic-context protein-binding microarrays (gcPBM) for model training and validation.

Main Results:

• Achieved a Pearson correlation of approximately 0.73 and a mean absolute error of 0.4 in predicting DNA binding affinities, outperforming conventional MMGBSA.
• Identified TF-DNA interfacial complementarity and hydrophobic interactions as key determinants of binding.
• Highlighted the need for further physical characterization of TF-DNA interfacial hydrogen bonding for sequence dependency.

Conclusions:

• The developed physics-informed ML framework offers a powerful approach for accurate and interpretable prediction of protein-DNA interactions.
• This method has the potential for scalable prediction of TF-DNA binding, advancing our understanding of gene regulation and disease.
• The findings pave the way for improved diagnostics and therapeutics targeting TF-DNA interactions.