Deficiencies in Molecular Dynamics Simulation-Based Prediction of Protein–DNA Binding Free Energy Landscapes

Transcriptional regulation allows cells to match their gene expression
profiles to their current requirements based on environment, cellular
physiological state, and extracellular signals. DNA binding transcription
factors are major agents of transcriptional regulation, and bind to
DNA with a factor-specific sequence preference to exert regulatory
effects. A crucial step in unraveling the logic of a regulatory network
is determining the sequence-specific binding affinity landscapes for
the transcription factors in it. While such landscapes can be measured
experimentally, the ability to predict them computationally would both
reduce the effort required to obtain the needed data and provide additional
insight into the key interactions shaping protein–DNA interactions.
Here we apply free energy calculations based on all-atom molecular
dynamics simulations to predict the changes in binding free energy
for all single base pair perturbations of the binding sites for four
eukaryotic transcription factors for which high-quality experimental
data exist. We find that the simulated results both vastly overestimate
the magnitude of changes in binding free energy, and frequently predict
the incorrect signs. These simulations will nevertheless serve as
a jumping-off point for refining our current representation of protein–DNA
interactions to allow quantitative reproduction of experimental data
on such systems in the future.