An edge-based formulation of elastic network models

Maxwell Hodges, Sophia N. Yaliraki, Mauricio BarahonabioRxiv 2019 · DOI

TL;DR

This paper introduces a novel "edge-based" formulation for Elastic Network Models (ENMs), shifting focus from node positions to the interactions (edges) between them. This new perspective naturally incorporates higher-order interactions and provides a more intuitive way to understand mechanical properties like "edge embeddedness" and network rigidity, with applications in protein dynamics.

What Problem Does This Paper Address?

Elastic Network Models (ENMs) are widely used to study the mechanical properties and dynamics of complex systems, particularly proteins, where full molecular dynamics simulations are computationally expensive. Traditional ENMs primarily focus on the positions of individual nodes (e.g., atoms or residues) and their collective motion. However, the interactions between these nodes, represented by springs or edges, are equally crucial for understanding how forces propagate and how the system responds to perturbations. The authors identify a gap in existing ENM frameworks by proposing an approach that directly analyzes these edge-level interactions, offering a more direct way to characterize mechanical responses and rigidity, especially when considering more complex interaction potentials beyond simple pairwise forces.

Methods

The core contribution is the development of an edge-based formulation for ENMs. Instead of focusing on node displacements ($u_i$), the model directly analyzes edge extensions ($e_\alpha$), which represent changes in the length of the "springs" connecting nodes.

Key methodological aspects include:

  • Geometric Incidence Matrix (B): This matrix relates node displacements to edge extensions ($e = B^T u$). Crucially, it incorporates the directional information of the edges in d-dimensional space, going beyond the standard graph incidence matrix.
  • Edge Mechanical Embeddedness ($\epsilon_\alpha$): A novel metric is introduced to quantify how much force an edge "feels" when a force is applied to the entire network. It's defined as $\epsilon_\alpha = 1 - g_\alpha T_{\alpha\alpha}$, where $g_\alpha$ is the spring constant and $T_{\alpha\alpha}$ relates to the network's response. A high $\epsilon_\alpha$ means the edge is well-integrated and dissipates forces effectively.
  • Inclusion of Higher-Order Interactions: The edge-based framework naturally accommodates the inclusion of three-center (angle) and four-center (dihedral) forces. This is achieved by deriving new stiffness matrices ($K_{angle}$, $K_{dihedral}$) that are added to the standard two-center bond stiffness matrix ($K_{bond}$) to form a total stiffness matrix ($K_{total}$).
  • Connection to Infinitesimal Rigidity: The authors demonstrate a direct relationship between their geometric incidence matrix and the rigidity matrix ($R$) of a structure, allowing for the analysis of rigid and flexible regions within the network.
  • Langevin Equation for Dynamics: For analyzing fluctuations, the standard Langevin equation is used, and the covariance matrix of edge fluctuations is derived in terms of the edge-based framework.

Datasets

The paper uses publicly available protein structures from the Protein Data Bank (PDB) for its applications. Specifically:

  • PDB code: 3ORZ (for the allosteric binding example, PDK1 protein).
  • PDB code: 4AKE (open conformation of ADK protein) and 1AKE (closed conformation of ADK protein) for residue-residue interaction analysis.

No new benchmarks or datasets were introduced.

Key Insights and Ideas

  • Shifting Perspective: The most significant insight is that framing ENMs in terms of "edge space" rather than "node space" offers a more direct and intuitive way to understand mechanical responses, especially for complex interactions.
  • "Edge Embeddedness" as a Meaningful Metric: The concept of edge mechanical embeddedness provides a quantifiable measure of how well an interaction is integrated into the network's mechanical response. This goes beyond simple connectivity or centrality measures and offers physical interpretation.
  • Natural Extension to Higher-Order Forces: The edge-based approach elegantly handles the inclusion of angle and dihedral constraints, which are crucial for accurately modeling molecular systems but can be cumbersome in traditional node-centric ENMs.
  • Rigidity and Strain Propagation: The direct link to infinitesimal rigidity highlights how the edge-based formulation can predict how strain propagates through a network, which is key for understanding phenomena like allostery in proteins.
  • Topological vs. Material Properties: The study on PDK1 suggests that while network topology (rigidity) plays a role, the specific "material properties" (spring constants) of the edges are also critical in determining how strain propagates, implying that allostery might not solely be a topological phenomenon.

Key Results

  • Allosteric Site Analysis (PDK1): The model identified that weak hydrophobic interactions near the allosteric site absorbed much of the applied force, potentially hindering efficient strain propagation to the active site. This effect was sensitive to the spring constants assigned to these interactions.
  • Residue-Residue Interactions (ADK): Analysis of ADK showed that the most significant edge fluctuations (residue-residue interactions) were concentrated in the lid and AMP-binding domains, regions known to undergo structural changes during the open-to-closed transition.
  • Exponential Decay of Response: For protein systems, the output edge extensions were found to decay exponentially with distance from the perturbation site, a behavior consistent with random networks.
  • Effect of Higher-Order Interactions: Including angle and dihedral constraints in the elastic network model significantly increased the network's rigidity, as visualized by infinitesimal rigidity analysis.

Limitations and Future Work

  • Spring Constant Assignment: The paper acknowledges the difficulty in assigning accurate spring constants for certain interactions (e.g., hydrophobic interactions) from standard molecular dynamics force fields. The results are sensitive to these assignments, and further work could explore more robust methods for parameterization.
  • Simplifications: The model relies on quadratic potential energy approximations and often uses simplified force constants. Real biological systems have more complex, non-linear potentials.
  • Future Work: The authors suggest that this edge-based framework is general and can be applied to other complex systems beyond proteins, such as glasses or materials. Further exploration of the "edge mechanical embeddedness" metric in different contexts is also a promising avenue. The connection to rigidity and the ability to include higher-order forces opens doors for more detailed mechanical analyses of complex structures.

Summary updated

Abstract

We present an edge-based framework for the study of geometric elastic network models to model mechanical interactions in physical systems. We use a formulation in the edge space, instead of the usual node-centric approach, to characterise edge fluctuations of geometric networks defined in d -dimensional space and define the edge mechanical embeddedness, an edge mechanical susceptibility measuring the force felt on each edge given a force applied on the whole system. We further show that this formulation can be directly related to the infinitesimal rigidity of the network, which additionally permits three- and four-centre forces to be included in the network description. We exemplify the approach in protein systems, at both the residue and atomistic levels of description.
BibTeX
@misc{hodges2019edge,
  author = {Maxwell Hodges and Sophia N. Yaliraki and Mauricio Barahona},
  title = {{An edge-based formulation of elastic network models}},
  year = {2019},
  doi = {10.1101/842732},
  howpublished = {bioRxiv},
}

Loading related content...

Figure