Abstract:
ANM-MC is a coarse-grained simulation technique, in which the anisotropic network model (ANM) generates the collective modes for deforming the protein structure towards target direction, and Monte Carlo (MC) algorithm minimizes the energy of the deformed structure. ANM-MC was modi ed in this thesis for the exploration of conformational transition pathways of large protein systems. The analyzed dataset consists of 13 large proteins that present either hinge-bending, DNA-binding or sheartype conformational transitions. All proteins contain multiple chains with 600 to 8000 residues in total. The distance (RMSD) between initial and nal conformations spans a broad range of 2.1 - 20.6 A. Initially, the adjustment of ANM-MC parameters to large proteins was performed using adenylate kinase and calmodulin. Later, detection of local implausible geometries for some large proteins lead to a modi ed version of ANM-MC with variable ANM deformations. As a result, all the proteins in the dataset approach the target structures successfully following suitable potential energy paths. Some of the proteins need more than the slowest 10 modes to approach the target better, speci cally 50 to 150 modes may be necessary. For such proteins, relatively lower indexed modes still play a dominant role during initial stages of the simulation with higher indexed modes chosen in later stages. Furthermore, coarse-grained intermediates along the transition pathway are reverse-mapped to full-atomistic structures by energy minimization. ANM-MC program is highly e cient based on the reported computational cost for each protein in the dataset.
