Docking-based long timescale simulation of cell-size protein systems at atomic resolution

Ilya A. Vakser, Sergei Grudinin, Nathan W. Jenkins, Petras J. Kundrotas, Eric J. Deeds

TL;DR

This paper introduces a novel computational method that combines protein docking with Monte Carlo simulations to achieve unprecedentedly long, all-atom resolution simulations of cell-sized protein systems. This breakthrough allows researchers to explore the dynamics of protein interactions within a cellular environment at a level of detail previously unattainable, paving the way for a deeper understanding of cellular mechanisms and disease.

What Problem Does This Paper Address?

Modeling the complex interactions of proteins within a living cell at an atomic level has been a significant challenge. Traditional molecular dynamics (MD) simulations, while providing atomic detail, are computationally too expensive to simulate the long timescales and large systems relevant to cellular processes. On the other hand, protein docking algorithms are efficient at predicting the static structures of protein complexes but lack the temporal dimension, failing to capture the dynamic nature of these interactions. This gap leaves a crucial area of cellular biology – the kinetics of protein association and dissociation in a crowded cellular environment – largely unexplored at a molecular level. The authors aim to bridge this gap by developing a method that leverages the efficiency of docking to enable long-timescale simulations of large protein systems.

Methods

The core of this work is a novel simulation protocol that combines two key ideas:

  • FFT-based Protein Docking: The Fast Fourier Transform (FFT) algorithm is used to efficiently precompute the intermolecular energy landscape for all possible pairwise protein interactions within the system. This precomputation maps the "docking poses" and their associated energies.
  • Monte Carlo (MC) Sampling with Minima Hopping: Instead of brute-force MD, the system's dynamics are sampled using a Monte Carlo approach. This MC protocol employs a "minima hopping" strategy, focusing on transitions between low-energy states (bound protein complexes) and avoiding high-energy, low-probability states. This is achieved by using the precomputed docking scores as the energy function.
  • Kinetic Propagation: The MC moves are designed to simulate the propagation of protein interactions in space and time. The protocol uses a Metropolis acceptance criterion, adapted to account for the detailed balance condition, to decide whether to accept a proposed move (a protein changing its binding partner).
  • Time Calibration: The time units of the simulation were calibrated against existing MD data for villin, establishing that one simulation step corresponds to approximately 20 nanoseconds.

This approach is novel in its application of efficient docking precomputation to drive a kinetic MC simulation, effectively allowing the system to explore a vast conformational space over extended periods.

Datasets

The study utilized existing protein structures from the Protein Data Bank (PDB) for their simulations. Specifically, they used:

  • Set 1 ("5 mix"): Five arbitrarily selected globular proteins of average size to represent a typical crowded cellular environment.
  • Set 2 ("GFP + 5 mix"): Set 1 plus Green Fluorescent Protein (GFP) to allow for comparison with experimental data on GFP diffusion.
  • Set 3 ("3 mix"): Three small proteins (ubiquitin, G protein B subunit, and villin) from a previous study.

No new datasets were generated or introduced as benchmarks; instead, the simulations were validated against existing experimental and MD simulation data.

Key Insights and Ideas

  • Bridging the Docking-Time Gap: The most significant insight is the successful integration of static protein docking with dynamic simulation. By using docking to define the energy landscape and MC to sample it, the method overcomes the temporal limitations of traditional docking.
  • Capturing Cellular Crowding Effects: The method is designed to explicitly model the crowded cellular environment, where transient protein interactions are dominant. This allows for a more realistic representation of cellular dynamics than simulations of dilute systems.
  • "Melting" Temperature as a System Property: The identification of a consistent "melting temperature" (T=100 in their arbitrary units) across different protein sets and volume fractions suggests that this parameter reflects the underlying energy landscape and the transition from a frozen, static state to a dynamic, interacting system. This provides a useful internal validation metric.
  • Scalability to Cell-Like Sizes: The computational efficiency of the approach allows for simulations of protein systems that approach the size of a cell, opening up new avenues for whole-cell modeling.

Key Results

  • Unprecedented Simulation Lengths: The method enables simulations reaching second-long trajectories for cell-sized protein systems at atomic resolution, a significant leap from the nanosecond to microsecond timescales typically achievable with all-atom MD.
  • Accurate Reproduction of Diffusion Rates: The simulations accurately recapitulated experimental data for GFP diffusion rates in a crowded environment (0.3 Ų/ns), aligning well with known values.
  • Realistic Aggregation Behavior: The predicted aggregation numbers (average number of proteins in a cluster) for the 5 mix set at physiological volume fraction (N=3.9) were in excellent agreement with experimental estimates for proteins like lysozyme and antibodies.
  • Diffusion Slowdown with Concentration and Size: The simulations correctly predicted that protein diffusion slows down significantly with increasing protein volume fraction and protein size, consistent with established experimental and theoretical observations.
  • Residence Times: The simulated protein residence times (around 570 ns) fell between experimental estimates and shorter MD predictions, suggesting a more accurate capture of transient interaction lifetimes.

Limitations and Future Work

  • Rigid-Body Approximation: The current model treats proteins as rigid bodies. The authors acknowledge that incorporating protein flexibility would be a crucial next step for more accurate modeling.
  • Simplified Force Field: The use of a step-function approximation for Lennard-Jones potentials and the reliance on GRAMM docking scores represent a simplification of more detailed all-atom force fields.
  • Limited Interaction Types: The current model focuses on protein-protein interactions. Future work will involve incorporating other macromolecules, membrane environments, and more sophisticated force fields that include electrostatics and solvent effects.
  • Move Set Optimization: The authors plan to further optimize the move set based on move acceptance probabilities to improve the simulation's efficiency and accuracy.
  • Computational Cost for Longer Trajectories: While significantly faster than MD, achieving multi-second simulations for cell-sized systems still requires substantial computational resources (estimated at about one year of CPU-core time for a 0.3- to 3-second simulation).

Despite these limitations, the study presents a powerful proof-of-concept that opens exciting possibilities for understanding complex biological processes at the molecular level.

Summary updated

Abstract

Computational methodologies are increasingly addressing modeling of the whole cell at the molecular level. Proteins and their interactions are the key component of cellular processes. Techniques for modeling protein interactions, thus far, have included protein docking and molecular simulation. The latter approaches account for the dynamics of the interactions but are relatively slow, if carried out at all-atom resolution, or are significantly coarse grained. Protein docking algorithms are far more efficient in sampling spatial coordinates. However, they do not account for the kinetics of the association (i.e., they do not involve the time coordinate). Our proof-of-concept study bridges the two modeling approaches, developing an approach that can reach unprecedented simulation timescales at all-atom resolution. The global intermolecular energy landscape of a large system of proteins was mapped by the pairwise fast Fourier transform docking and sampled in space and time by Monte Carlo simulations. The simulation protocol was parametrized on existing data and validated on a number of observations from experiments and molecular dynamics simulations. The simulation protocol performed consistently across very different systems of proteins at different protein concentrations. It recapitulated data on the previously observed protein diffusion rates and aggregation. The speed of calculation allows reaching second-long trajectories of protein systems that approach the size of the cells, at atomic resolution.
BibTeX
@inproceedings{vakser2022docking,
  author = {Ilya A. Vakser and Sergei Grudinin and Nathan W. Jenkins and Petras J. Kundrotas and Eric J. Deeds},
  title = {{Docking-based long timescale simulation of cell-size protein systems at atomic resolution}},
  booktitle = {Proceedings of the National Academy of Sciences},
  year = {2022},
  doi = {10.1073/pnas.2210249119},
}
Figure