Surveying biomolecular frustration at atomic resolution

Mingchen Chen, Xun Chen, Nicholas P. Schafer, Cecilia Clementi, Elizabeth A. Komives, Diego U. Ferreiro, Peter G. WolynesNature Communications 2020 · DOI

TL;DR

This paper introduces a new computational framework to precisely map "frustration" – conflicting interactions – within biomolecules at atomic resolution. This method reveals that frustrated regions are not random but strategically located, correlating with key functional sites in proteins and, importantly, with the specificity of drug-protein interactions, offering a powerful new tool for drug discovery.

What Problem Does This Paper Address?

Biomolecules, like proteins, need to be both stable enough to maintain their structure and flexible enough to perform their functions. This delicate balance is achieved through a complex interplay of interactions. Some interactions are "minimally frustrated," contributing to a stable, "funneled" energy landscape that guides the molecule to its functional state. However, other interactions are "frustrated," meaning they conflict with each other. These frustrated regions are crucial for allowing the molecule to move, interact with other molecules, and adapt to different environments.

While previous methods could identify frustration at a coarse-grained level, precisely locating and quantifying these frustrated interactions at the atomic scale, especially in complex scenarios like protein-ligand binding, remained a challenge. This lack of atomic resolution limited our understanding of how subtle differences in molecular interactions contribute to biological specificity and how drugs achieve their targeted effects. The paper aims to bridge this gap by providing a computationally efficient and high-resolution method for analyzing biomolecular frustration.

Methods

The core of this work is an atomistic frustration analysis framework. This framework quantifies "frustration" by examining the statistics of energy changes when the local environment of a specific site within a biomolecule is altered.

Here's a breakdown of the key methodological aspects:

  • Decoy Generation: For each contact (interaction between two residues) in a protein, the authors generate a set of "decoys." This involves randomly shuffling the amino acid identities and/or their positions within the local environment. This process simulates alternative configurations.
  • Energy Calculation: The Rosetta all-atom force field (specifically the REF2015 version) is used to calculate the energy of both the native contact and the generated decoys.
  • Frustration Index: A "frustration index" is calculated for each contact. This index is essentially a Z-score that measures how favorable the native contact's energy is compared to the energies of the decoy contacts. A high Z-score indicates a stable, minimally frustrated interaction, while a low Z-score suggests a frustrated interaction.
  • Efficiency Improvement: A key innovation is the improved efficiency of decoy generation compared to previous atomistic methods. Instead of relying solely on computationally expensive molecular dynamics simulations, they employ a faster approach of randomizing residue identities and repacking side chains onto a fixed backbone. This significantly speeds up the calculations, enabling a large-scale survey.
  • Handling Repulsive Forces: To avoid artificial inflation of frustration due to harsh repulsive forces in the force field, they separate repulsive and attractive terms, focusing on the latter for a more realistic assessment of functional frustration.

Datasets

The study utilizes several publicly available datasets:

  • Nonredundant Set of Monomeric Protein Domains: A set of 314 protein domains with high-resolution structures was used to analyze the spatial distribution of frustration in individual proteins.
  • Catalytic Site Atlas: Entries from this atlas were used to investigate frustration patterns around enzyme active sites.
  • Database of Homologous Proteins in Different Conformational States: This dataset was used to study frustration in allosteric proteins.
  • Database of Protein Assemblies: Used to analyze frustration at protein-protein binding interfaces.
  • EGFR-kinases in Complex with Inhibitors: A dataset of over 700 proteins complexed with ligands, specifically focusing on EGFR-kinases and their inhibitors, was used to correlate binding affinity with frustration.
  • COX-1 and COX-2 Inhibitor Complexes: A dataset of 54 COX-inhibitors in complex with COX-1 and COX-2 was used to assess drug specificity.

Key Insights and Ideas

  • Strategic Frustration: Frustration isn't just noise; it's a precisely evolved feature. Highly frustrated interactions are not randomly distributed but tend to cluster at specific sites, often on protein surfaces, correlating with regions involved in function, motion, and interaction.
  • Funneled Landscapes Extend to Ligand Binding: The concept of a "funneled" energy landscape, crucial for protein folding, also applies to how ligands bind to proteins. A smooth, minimally frustrated binding pocket leads to a more specific and effective interaction.
  • Atomic Resolution Matters for Specificity: While coarse-grained models provide insights, atomic resolution is essential for truly understanding the nuances of molecular recognition, particularly for distinguishing between similar binding sites or targets.
  • Drug Specificity is Linked to Minimal Frustration: The study provides strong evidence that drugs with higher specificity exhibit minimally frustrated binding pockets. This suggests that designing drugs with "smoother" binding landscapes could lead to more targeted therapies with fewer side effects.
  • Water's Role in Frustration: The paper highlights differences in how atomistic models (like Rosetta) and coarse-grained models (like AWSEM) handle water-mediated interactions. This suggests that accurately modeling "wet" interfaces, common in protein complexes, may require further refinement in force fields.

Key Results

  • Localized Frustration Correlates with Function: Analysis of monomeric proteins showed that highly frustrated contacts are enriched on surfaces, often near functionally important regions.
  • Allosteric Proteins Show Frustrated Reconfiguration Regions: In allosteric proteins, regions undergoing conformational changes were found to be enriched in highly frustrated interactions.
  • Enzyme Active Sites Have Specific Frustration Patterns: Frustration analysis around enzyme active sites revealed clustering of frustrated interactions, potentially indicating functional roles.
  • Dry Interfaces are Minimally Frustrated: Protein-protein complexes with "dry" binding interfaces (not heavily mediated by water) showed largely minimally frustrated interactions.
  • Ligand Binding Reduces Frustration: For many protein-ligand complexes, binding of the ligand significantly reduced frustration within the binding pocket, creating a more stable environment.
  • Correlation Between Binding Affinity and Frustration: In EGFR-kinase inhibitors, a correlation was observed between the number of minimally frustrated interactions in the binding pocket and the measured binding affinity.
  • Selective Drugs Have Distinct Frustration Signatures: Selective COX-2 inhibitors showed a greater increase in minimally frustrated interactions when binding to COX-2 compared to COX-1, distinguishing them from non-selective inhibitors.

Limitations and Future Work

  • Force Field Dependence: The accuracy of the frustration analysis is dependent on the underlying atomistic force field (Rosetta in this case). The paper notes potential limitations in how Rosetta handles water-mediated interactions compared to other models like AWSEM.
  • Computational Cost: While improved, atomistic frustration analysis is still computationally intensive, especially for very large systems.
  • Future Directions:
    • Improved Force Fields: Further development of force fields, particularly those that better account for water-mediated interactions, is crucial for accurate analysis of "wet" interfaces.
  • Broader Application: The framework can be readily adapted to other atomistic force fields and applied to a wider range of biomolecular systems.
  • Drug Discovery Pipeline: Integrating this frustration analysis into drug discovery pipelines could accelerate the screening of potential drug candidates for improved specificity and reduced off-target effects.
  • Understanding Dynamics: Future work could explore how frustration profiles evolve dynamically and how they relate to protein dynamics beyond static structures.

Summary updated

Abstract

To function, biomolecules require sufficient specificity of interaction as well as stability to live in the cell while still being able to move. Thermodynamic stability of only a limited number of specific structures is important so as to prevent promiscuous interactions. The individual interactions in proteins, therefore, have evolved collectively to give funneled minimally frustrated landscapes but some strategic parts of biomolecular sequences located at specific sites in the structure have been selected to be frustrated in order to allow both motion and interaction with partners. We describe a framework efficiently to quantify and localize biomolecular frustration at atomic resolution by examining the statistics of the energy changes that occur when the local environment of a site is changed. The location of patches of highly frustrated interactions correlates with key biological locations needed for physiological function. At atomic resolution, it becomes possible to extend frustration analysis to protein-ligand complexes. At this resolution one sees that drug specificity is correlated with there being a minimally frustrated binding pocket leading to a funneled binding landscape. Atomistic frustration analysis provides a route for screening for more specific compounds for drug discovery.
BibTeX
@inproceedings{chen2020surveying,
  author = {Mingchen Chen and Xun Chen and Nicholas P. Schafer and Cecilia Clementi and Elizabeth A. Komives and Diego U. Ferreiro and Peter G. Wolynes},
  title = {{Surveying biomolecular frustration at atomic resolution}},
  booktitle = {Nature Communications},
  year = {2020},
  doi = {10.1038/s41467-020-19560-9},
}
Figure