A local minimum is relatively easy to find while the global minimum is computationally intractable in general.

With rahul ravindrudu, Vlad Sukhoy,  ...

In principle, the protein native structure can be located at a structure that minimizes the protein’s total potential energy. However, assuming that the energy function is well defined and available, the problem of minimizing the function is still a big challenge, for it requires a global minimum in the whole protein conformation space. The solution to a global optimization problem is difficult unless a special algorithm that can exploit the special structure of the objective function and make the global search real effective. Great progress has been made towards the solution of the global optimization problem associated with protein folding, yet the problem remains as a grand challenge of modern science.

Phone: 515-294-8165

Fax: 515-294-5454

E-mail: zhijun@iastate.edu

Contact information:

We are developing a global energy minimization algorithm by utilizing several advanced optimization techniques including trust-region local optimization, energy landscape smoothing, and normal modes guided global search.

 

A global optimization algorithm can be implemented to be scalable on parallel machines so that the required computation for locating a global minimum, even if it is unusually demanding, can nevertheless be shared effectively among available computing resources.

 

We are implementing our global optimization algorithms on various different parallel architectures. Our ultimate goal is to develop a parallel energy minimization code that can effectively fold a protein “from scratch”.