Molecular modeling of proteins and DNA is an attractive goal because it allows to gain insight into dynamic behavior of molecules on atomistic level. Such studies have a great potential to complement existing experimental techniques in investigating mechanisms of biomolecular phenomena. However, due to large size and ruggedness of free energy landscapes of biopolymers, simulations of long-time scale events often suffer from the pseudoergodicity problem, which manifests as inability to explore configurational space of interest within available computation time. The studies presented here reflect efforts to resolve this problem both by exploring advantages and limitations of the existing simulation techniques in studying behavior of specific biomolecular systems, and by testing new simulation methodologies.

Part I is concerned with the mechanism of 8-oxoguanine DNA lesion recognition by the bacterial DNA repair enzyme MutM. Two qualitative studies using the Targeted Molecular Dynamics technique explore possible molecular interactions in the MutM/DNA complex associated with the enzyme’s sliding along the ds-DNA and the extrusion of the interrogated base into the MutM’s catalytic pocket. The findings suggest that MutM may require rocking motion of the bases encountered in its sliding along the DNA. The rocking motion of the oxoguanine is likely to be restricted due to repulsive electrostatic interactions of its O8 atom with the DNA phosphate backbone and may result in braking of the sliding motion of the MutM. This led to the proposal of the braking recognition mechanism in which recognition occurs due to arrest of the sliding process at the lesion site, which makes possible the otherwise slower process of extruding the base into the catalytic pocket of the MutM for excision. The second study suggests that binding between the conserved Arg 112 residue of MutM and the cytosine estranged during the oxoguanine extrusion may be important to prevent partial oxoguanine extrusion from becoming an alternative sliding pathway.

Chapter 1.4 describes an attempt to investigate a possibility of long-range lesion recognition by MutM by computing free energy of MutM/DNA complexes. The Orthogonal Space Random Walk technique used for the free energy calculations in this experiment represents one of recent advancements in the generalized ensemble simulation methodology. Failure to obtain converged free energy estimates for the MutM/DNA complexes led to the discussion of possible limitations of this technique and to the proposal that MutM binding to the DNA in the vicinity of a lesion may require a global conformational reorganization of the DNA in comparison to lesion-free MutM/DNA complexes.

Part II presents three small model systems studies of the efficiency of generalized ensemble simulation techniques and reflects a part of the recent methodological development in this field. The generalized ensemble techniques utilize sampling of the configrational space with modified probabilities in order to overcome the barrier crossing problem. Chapters 2.2 and 2.4 are concerned with history-dependant methods of formulating a priory unknown efficient probability modifications. In these studies the use of the Wang-Landau recursion approach in the metadynamics technique and the hybrid Wang-Landau recursion/adaptive reweighing approach for the Simulated Scaling technique are tested. The chapter 2.3 is concerned with an attempt of resolving the diffusion sampling problem associated with the generalized ensemble methods by implementing the Self-Guided Langevin dynamics approach for the Essential Energy Space modality of the metadynamics technique. All three studies demonstrated superior sampling efficiency of the proposed method enhancements for the selected model systems.