Residue Geometry Networks: A Rigidity-Based Approach to the Amino Acid Network and Evolutionary Rate Analysis.

Alexander S Fokas,Alex W Chin,Daniel J Cole,Sebastian E Ahnert

doi:10.1038/srep33213

Abstract

Amino acid networks (AANs) abstract the protein structure by recording the amino acid contacts and can provide insight into protein function. Herein, we describe a novel AAN construction technique that employs the rigidity analysis tool, FIRST, to build the AAN, which we refer to as the residue geometry network (RGN). We show that this new construction can be combined with network theory methods to include the effects of allowed conformal motions and local chemical environments. Importantly, this is done without costly molecular dynamics simulations required by other AAN-related methods, which allows us to analyse large proteins and/or data sets. We have calculated the centrality of the residues belonging to 795 proteins. The results display a strong, negative correlation between residue centrality and the evolutionary rate. Furthermore, among residues with high closeness, those with low degree were particularly strongly conserved. Random walk simulations using the RGN were also successful in identifying allosteric residues in proteins involved in GPCR signalling. The dynamic function of these residues largely remain hidden in the traditional distance-cutoff construction technique. Despite being constructed from only the crystal structure, the results in this paper suggests that the RGN can identify residues that fulfil a dynamical function.

Highlights

Amino acid networks (AANs) abstract the protein structure by recording the amino acid contacts and can provide insight into protein function
Dynamical techniques derive edge information from molecular dynamics (MD) simulations in which edges are introduced based on the percentage of conformations in which two residues are in contact during a simulation[7,8]
Using the weighted residue geometry network (RGN), we find the same trend, as well as an increase in the weighted betweenness centrality correlation when compared to the correlation measured using the unweighted RGN (unRGN)

Summary

Introduction

Amino acid networks (AANs) abstract the protein structure by recording the amino acid contacts and can provide insight into protein function. We describe a novel AAN construction technique that employs the rigidity analysis tool, FIRST, to build the AAN, which we refer to as the residue geometry network (RGN) We show that this new construction can be combined with network theory methods to include the effects of allowed conformal motions and local chemical environments. AANs were constructed using a physical distance-cutoff (DC)[2], whereby edges are placed between residues that are within a certain DC This method showed that AANs display small world properties, where few nodes are direct neighbours, but most nodes can be reached in few steps. Properties of such networks, including average degree and clustering coefficient, have been employed to score and subsequently discriminate between native and non-native structures[3] While insightful, such AANs are considered coarse grained methods, as they only store information concerning the general protein shape. A computationally cheap approach that uses only the static structure and maintains a comparable level of validity would be profitable for certain applications, in particular, for the de novo design of protein function

Methods

Results

Conclusion