From Quantum Chemistry to Networks in Biology: A Graph Spectral Approach to Protein Structure Analyses

TL;DR

This paper introduces a graph spectral method integrating quantum chemistry and network theory to analyze protein structures, revealing key residues and structural features with applications in understanding biological functions and diseases.

Contribution

It presents a novel multidisciplinary approach combining spectral graph theory and biological insights for detailed protein network analysis.

Findings

Effective identification of crucial amino acids for structural integrity.

Enhanced protein structure comparison through spectral analysis.

Insights into allosteric effects via node clustering changes.

Abstract

In this perspective article, we present a multidisciplinary approach for characterizing protein structure networks. We first place our approach in its historical context and describe the manner in which it synthesizes concepts from quantum chemistry, biology of polymer conformations, matrix mathematics, and percolation theory. We then explicitly provide the method for constructing the protein structure network in terms of non-covalently interacting amino acid side chains and show how a mine of information can be obtained from the graph spectra of these networks. Employing suitable mathematical approaches, such as the use of a weighted, Laplacian matrix to generate the spectra, enables us to develop rigorous methods for network comparison and to identify crucial nodes responsible for the network integrity through a perturbation approach. Our scoring methods have several applications in structural biology that are elusive to conventional methods of analyses. Here, we discuss the instances of: (a) Protein structure comparison that include the details of side chain connectivity, (b) The contribution to node clustering as a function of bound ligand, explaining the global effect of local changes in phenomena such as allostery and (c) The identification of crucial amino acids for structural integrity, derived purely from the spectra of the graph. We demonstrate how our method enables us to obtain valuable information on key proteins involved in cellular functions and diseases such as GPCR and HIV protease, and discuss the biological implications. We then briefly describe how concepts from percolation theory further augment our analyses. In our concluding perspective for future developments, we suggest a further unifying approach to protein structure analyses and a judicious choice of questions to employ our methods for larger, more complex networks, such as metabolic and disease networks.