Chinwe Ekenna, Shawna Thomas, and Nancy Amato,
RSS Workshop on Robotics Methods for Structural and Dynamic Modeling of Molecular Systems, Berkeley, California, July 2014.
Abstract: Robotic motion planning algorithms such as Probabilistic Roadmap Methods (PRMs) have been successful in simulating the protein folding process by building a roadmap, or model, of the folding landscape. This roadmap is constructed by sampling protein conformations and connecting them together with energetically feasible transitions. In this work, we propose an adaptive method to dynamically select an appropriate connection method from a set of connection method candidates. Our framework, Adaptive Neighbor Connection (ANC), learns which strategy to use by examining their success and cost over time. Thus, it frees the user of the burden of selecting the best strategy and allows this selection to change over time. We compare ANC to 6 other distance-based connection methods on a set of 7 well- studied proteins. We show that ANC builds roadmaps quickly with high quality folding pathways.
Research Pages: