Nonparametric Structure Models in Local Protein Conformation Sampling and Design

Abstract

Protein design relies of the identification of a sequence that specifically encodes a target conformation as a folded native state. This native states is encoded by a combination of energetically favorable local and nonlocal structural features. Rapid identification and design of conserved, local structural features may be used to enable and accelerate design of functional or novel non-local interactions. The work here describes the implementation and initial analysis of a turn design strategy based on nearest-neighbor queries against an extensible database of known protein structures. We outline the implementation of this database and search system within the Rosetta biomolecular modeling framework, the successful application of this approach to the atomic-level design of helical hairpin turns and the extension of this approach to combinatorial diversification of a ligand-binding scaffold via a distributed simulation integrating the Rosetta and PyData software stacks.