sing Experimental Protein Data to Construct a Configuration Space for Protein Folding Simulation Using Motion Planning Tech.

Document Type



Hans Dulimarta, dulimarh@gvsu.edu

Embargo Period



Deriving the structure of a protein from only its DNA sequence is theoretically possible, but the computational demands are so enormous that it’s impossible to complete a simulation of any average-sized protein in one’s lifetime. There is, therefore, a great deal of effort being expended on techniques that can shorten the time it takes to do a simulation. One technique being explored is the application of Probabilistic Road Map techniques, which were developed to solve the problem of moving a robot arm with multiple degrees of freedom from one configuration to another, to protein folding, using the insight that a protein is a chain of amino acids with fixed-length links between them. To date, the technique has been used to explore known proteins, starting from a known configuration, varying it and seeing how quickly and closely the simulation brings you back to your starting point. Unfortunately, this is not terribly useful in the case of proteins with unknown structures, but finding a way to constrain the search space is difficult. A simple method is to use aggregate data about known protein structures to construct a search space. That search space imposes some geometric constraints on the possible paths and opens up several possible approaches. The most interesting is the possibility of doing multiple passes within the search space, with each one being more granular and based on the results of the prior passes. To do these experiments, a code base from a prior experiment by Apaydin, et al (2002) is being modified so that the experimental data and multiple passes can both be taken into account.

This document is currently not available here.