Algorithms For Computational Description Of Large Scale Conformational Transitions Of Proteins
Conformational transitions of protein macromolecules are key elements in controlling functionality of proteins by changing structural and functional properties of protein molecules. These transitions consist of structural adjustments at spatial scales of 10-100 Å, between one to two orders of magnitude larger than a typical interatomic distance (2 Å). The difference in temporal scales of atomic motions and conformational transitions spans an even larger range. The transition time, microseconds to milliseconds, is between six to twelve orders of magnitudes larger than typical atomic oscillations (femtoseconds to picoseconds). The atomic resolution of studied systems (10 - 100 thousands of atoms) and long range inter-atomic interactions dictate the computational cost of simulations. This dissertation discusses several strategies to overcome these scaling issues in computational studies of conformational transition: the temporal, spatial, and size scales. The presented algorithms provide thermodynamics, kinetics and structural descriptions of conformational transitions at overall computational costs several orders of magnitude lower than the straight forward Molecular Dynamics approach. The presented algorithms are based on combination of coarse-graining strategies of (i) boundary value approach by an action minimization, (ii) statistical coarse-grained potentials, and (iii) Milestoning algorithm with an extension to complex (nonlinear) reactions. All presented algorithms are implemented in MOIL molecular modeling package and are parallelized to run effectively on high performance computing clusters.
dissertation or thesis