Generalized correlation-based dynamical network analysis: a new high-performance approach for identifying allosteric communications in molecular dynamics trajectories

Marcelo C. R. Melo, Rafael C. Bernardi, César de la Fuente‐Núñez, Zaida Luthey‐Schulten

Abstract

Molecular interactions are essential for regulation of cellular processes, from the formation of multiprotein complexes, to the allosteric activation of enzymes. Identifying the essential residues and molecular features that regulate such interactions is paramount for understanding the biochemical process in question, allowing for suppression of a reaction through drug interventions, or optimization of a chemical process using bioengineered molecules. In order to identify important residues and information pathways within molecular complexes, the Dynamical Network Analysis method was developed and has since been broadly applied in the literature. However, in the dawn of exascale computing, this method is generally limited to relatively small biomolecular systems. In this work we provide an evolution of the method, application and interface. All data processing and analysis is conducted through Jupyter notebooks, providing automatic detection of important solvent and ion residues, an optimized and parallel generalized correlation implementation that is linear with respect to the number of nodes in the system, and subsequent community clustering, calculation of betweenness of contacts, and determination optimal paths. Using the popular visualization program VMD, high-quality renderings of the networks over the biomolecular structures can be produced. Our new implementation was employed to investigate three different systems, with up to 2.5 M atoms, namely the OMP-decarboxylase, the Leucyl-tRNA synthetase complexed with its cognate tRNA and adenylate, and the respiratory complex I in a membrane environment. Our enhanced and updated protocol provides the community with an intuitive and interactive interface, which can be easily applied to large macromolecular complexes.

Introduction

In structural biology, allostery describes the mechanism by which a protein senses a signal at one site and transmits a functional change to a distant site. While Molecular Dynamics (MD) simulations are the primary tool for observing these processes, they generate massive amounts of data that are difficult to parse. Traditional Dynamical Network Analysis (DNA) maps atoms to "nodes" and their interactions to "edges," but the computational cost typically scales quadratically ($O(N^2)$) with the number of nodes. As MD simulations enter the exascale era—modeling complexes with millions of atoms—standard DNA tools have become a significant bottleneck. This paper introduces a high-performance evolution of DNA that scales linearly, enabling the identification of allosteric pathways in massive molecular systems.

Methodology: Achieving Linear Scaling

The authors developed a new implementation of DNA that leverages the inherent sparsity of molecular systems. Key technical advancements include:

  • Contact-Based Sparsity: Instead of calculating generalized correlation coefficients ($r_{MI}$) for every possible pair of atoms, the tool generates a contact matrix and only calculates correlations for nodes within a 4.5Å distance. This shifts complexity from $O(N^2)$ to $O(N)$.
  • Generalized Correlation: Unlike standard Pearson coefficients, which only capture linear motion, this tool uses Shannon mutual information to capture non-linear and perpendicular atomic correlations.
  • Modern Tech Stack: Implemented as a Python package using Numba and Cython for high-performance math, the tool is integrated into Jupyter Notebooks for interactivity and features a new GUI for VMD (Visual Molecular Dynamics) for 3D renderings.
  • Graph Algorithms: The software utilizes the Floyd-Warshall algorithm for finding optimal communication paths and Louvain heuristics for community clustering.

Validation Across Biological Scales

The tool was validated across three biological systems of increasing complexity, ranging from small enzymes to massive membrane-embedded complexes.

The three benchmark systems: (a) OMP-decarboxylase, (b) Leucyl-tRNA synthetase, and (c) Respiratory Complex I in a membrane environment.
  1. OMP-decarboxylase: A small, efficient enzyme (~48,000 atoms) used as a baseline.
  2. Leucyl-tRNA synthetase (LeuRS) complex: A protein-tRNA-ligand complex (~260,000 atoms) used to identify known identity elements.
  3. Respiratory Complex I: A massive transmembrane protein (~2,500,000 atoms) used to test the limits of the software in a lipid membrane environment.

Key Insights and Results

The Sparsity Advantage

The authors found that in a typical simulation, only about 0.7% of possible node pairs are in contact. By ignoring the other 99.3%, they achieved massive speedups without losing biological accuracy. For a ribosome system, analysis time dropped from 12 hours to 3 minutes; for an HIV capsid simulation, it plummeted from 4 days to 8 minutes.

Identifying Allosteric Hubs

By calculating "betweenness centrality," the tool identifies specific residues that act as information bottlenecks. In the LeuRS complex, the analysis identified a specific Lysine-to-RNA relay (Lys619 to G71) critical for stabilizing the complex and recognized experimentally verified "identity elements" essential for tRNA recognition.

Detailed view of the Leucyl-tRNA synthetase complexed with its cognate tRNA and adenylate.

Dynamic Allostery and Lipids

The tool allows trajectories to be analyzed in "windows," revealing that allosteric pathways are dynamic and can evolve over time. Furthermore, the study of Complex I demonstrated that lipids in the membrane can be integrated into the dynamical network, showing how the membrane itself influences protein communication.

Conclusion and Future Work

While the correlation calculation is now linear, the initial contact matrix determination still involves an $O(N^2)$ distance check. Future work may mitigate this by using advanced neighbor-searching algorithms. Additionally, while the tool provides the "what" and "where" of communication, determining the optimal temporal window size for analysis still requires expert judgment to ensure fluctuations are captured accurately.

Summary updated

Abstract

Molecular interactions are essential for regulation of cellular processes from the formation of multi-protein complexes to the allosteric activation of enzymes. Identifying the essential residues and molecular features that regulate such interactions is paramount for understanding the biochemical process in question, allowing for suppression of a reaction through drug interventions or optimization of a chemical process using bioengineered molecules. In order to identify important residues and information pathways within molecular complexes, the dynamical network analysis method was developed and has since been broadly applied in the literature. However, in the dawn of exascale computing, this method is frequently limited to relatively small biomolecular systems. In this work, we provide an evolution of the method, application, and interface. All data processing and analysis are conducted through Jupyter notebooks, providing automatic detection of important solvent and ion residues, an optimized and parallel generalized correlation implementation that is linear with respect to the number of nodes in the system, and subsequent community clustering, calculation of betweenness of contacts, and determination of optimal paths. Using the popular visualization program visual molecular dynamics (VMD), high-quality renderings of the networks over the biomolecular structures can be produced. Our new implementation was employed to investigate three different systems, with up to 2.5M atoms, namely, the OMP-decarboxylase, the leucyl-tRNA synthetase complexed with its cognate tRNA and adenylate, and respiratory complex I in a membrane environment. Our enhanced and updated protocol provides the community with an intuitive and interactive interface, which can be easily applied to large macromolecular complexes.
BibTeX
@article{melo2020generalized,
  author = {Marcelo C. R. Melo and Rafael C. Bernardi and César de la Fuente‐Núñez and Zaida Luthey‐Schulten},
  title = {{Generalized correlation-based dynamical network analysis: a new high-performance approach for identifying allosteric communications in molecular dynamics trajectories}},
  journal = {The Journal of Chemical Physics},
  year = {2020},
  doi = {10.1063/5.0018980},
}
Figure