Simulations of nucleic acids under stress, in solution, and complexed to proteins

Molecular dynamics (MD) simulations have become an important tool in advancing our understanding of the structure, function, and dynamics of biomolecules. It is only within the last several years that computational resources and techniques have advanced enough to allow for simulations extending into the multiple nanosecond range, systems of multiple biomolecules, and accurate free energy calculations. This dissertation reports on multiple studies focused on MD simulations of nucleic acids. In the first the effects of torque and tension on canonical B-DNA and A-RNA helices are examined and transitions to P-form, supercoiled P-form, and denatured states are observed. The free energies and forces of the transitions to P-form are then computed, which in the case of B-DNA agree with an experimentally derived torque-tension phase diagram while with A-RNA offer predictions for forces and torques required in single molecule experiments. Following this a study of the entropies of left and right handed DNA and RNA duplexes is presented aimed at understanding the effects of sequence, solvent, and ionic conditions on experimentally observed thermally induced transitions. Third, an in depth analysis of the free energies and mechanisms of DNA supercoil relaxation by human topoisomerase I is presented. It demonstrates the possibility of distinct mechanisms (with similar rates) for the relaxation of positive and negative supercoils while suggesting the presence of "semi-open" states. Additional calculations with the inhibitor topotecan show distinct mechanistic differences which selectively inhibit the rate of positive supercoil relaxation (in accord with experimental results). Finally a new method for enhanced sampling of rare events is developed in which a negative frictional coefficient is utilized in Langevin dynamics to introduce energy into the system along a specified reaction coordinate. It is demonstrated that this method efficiently scales to systems of many dimensions and with proper reweighting correlation functions for the physical positive friction system may be recovered from these negative friction calculations.