Potentials of mean force as a starting point for understanding biomolecular interactions

Computer simulations on the molecular dynamics of biological molecules have become commonplace in the field of biophysics, since they can be used to explore the behavior of large molecules with a level of detail not possible in experiments. It is necessary to be cautious, however, when designing and interpreting such simulations, as the techniques commonly used to improve efficiency of simulations can lead to unrealistic results. The work presented in this dissertation explores ways in which the accuracy of molecular dynamics simulations can be both improved and validated by experimental data, primarily through the use of potentials of mean force (PMFs). Experimental input was used in the development of an umbrella sampling protocol by fitting restraining potentials to an experimental PMF. The method was tested on a model peptide using a "guiding" PMF from simulations and then validated using an experimental PMF from force manipulation studies on a mechanical protein. The results show that the experimentally guided umbrella sampling replicates the appropriate pathways for both systems, whereas naively chosen potentials fail to do so. Experimental findings were also used in the design of steered molecular dynamics simulations on the beta domain of streptokinase. High-temperature simulations were used to smooth the energy surface and enable the system to explore alternate unfolding pathways. The results show three distinct pathways, in agreement with experimental evidence of three types of behavior under force. The simulations reveal that the source of the differences are hydrophobic interaction in the core of the protein. Multi-dimensional PMFs were calculated to describe these pathways energetically. All-atom simulations were also used to study a different type of system, the interactions between DNA and a Polyamidoamine dendrimer. Both the dendrimer and DNA were found to deform substantially upon binding. While the interactions were shown to be driven primarily by electrostatics, we also find that ordered waters extend the interaction distance beyond the range of direct electrostatics for one orientation of the dendrimer. These water effects contribute almost a third of the total interaction free energy of the system. PMFs calculated in these simulations were used to calculate force extension curves which agree with experiments on DNA condensed by dendrimers.