Free energy is a fundamental physical quantity in considerations of chemical reactions and biomolecular processes, determining the spontaneity of processes and stability of intermediates. Computational techniques based on statistical mechanical principles offer powerful means of obtaining free energy of molecular processes difficult to address directly in experiments.

This thesis explores techniques useful for the determination of free energies of biomolecular conformational equilibria by computational methods using molecular dynamics simulations. A simplest approach for obtaining the free energy is to use unbiased molecular dynamics trajectories to find probability densities. Limitations of such unbiased sampling approaches are illustrated by a model system of polyproline peptide.

Molecular dynamics simulations are limited to periods of ten or so nanoseconds to microseconds, which is inadequate to sample rugged free energy landscapes. Various biased sampling methods exist, enhancing the efficiency of sampling by imposing known constraining potentials, making the exploration of otherwise unfavorable configurations favored. We applied one technique, umbrella sampling, to the large scale conformational change of myosin.

Another method, self healing umbrella sampling, was also tested, with the model system of dialanine, comparing it to other established techniques.