Assembly of low-energy protein conformations with heterogeneous fragments

The most successful ab-initio protein structure prediction methods employ fragment-based assembly. This technique is usually implemented in the context of a Metropolis Monte Carlo template. Fragment length is an important consideration, as it controls the complexity of the discretization of the search space and the associated energy surface. This work measures the impact of employing multiple, heterogeneous fragment lengths in the context of a multi-stage probabilistic search framework.