Developing parallel programs and algorithms for protein models containing different types of complexity

Understanding the physical behaviors of proteins, at present, is one of the most active research topics. Although modeling efforts in this field has provided some encouraging results, a computational model capable of explaining the folding mechanism of proteins was not developed yet. One of the main reasons behind this failure is that the force fields are not good enough for modeling proteins, and limitation of computational power. To get full performance from today's computers can be achieved by using parallel programming algorithms instead of traditional serial programming. Therefore, it is critically important to develop new protein models, which can run with parallel algorithms. In this dissertation study, a coarse-grained protein model with explicit chain representation, which also relies on crystal structure information has been formed. Then, by taking into account the molecular dynamic simulation technique and by using CUDA programming language, new parallel codes were developed for this protein model. Thermodynamic and kinetic behaviors of 15 different proteins were investigated with these programs. Our results indicate that the main parameter, which determines protein folding thermodynamics and kinetics, is the average contact amount per amino asit, or, average connectivity.