SUMMARY

The native conformation (i.e., natural shape) of a protein largely determines its function, and determining it is known as the "protein-folding problem." Two typical approaches to this problem are molecular dynamic (MD) simulation and optimization. MD simulations are limited to smaller proteins because of significant computation demands. Optimization approaches, searching the conformation space, require less computation, but a global solution is difficult. The proposed research captures benefits of both MD and optimization approaches using a controls approach. A controls approach is considered because a protein can be modeled as a chain of rigid links connected by joints. This is similar to a serial robotic manipulator so related algorithms apply. For example, an existing simple optimization technique is actually equivalent to a proportional control algorithm to minimize joint torques due to molecular potentials. In preliminary simulations, we have applied an optimal control algorithm used in visual servoing to yield a proportional control with plant estimation, given by the second derivative of the potential energy. Our method yields better results than the simple proportional control and demonstrates the promise of approaching protein-folding as an optimal control problem. This research will explore suitable applications of classical, optimal, and adaptive controls as candidate algorithms. Important properties used to quantify results include stability, convergence, and conformation accuracy.