Computational strategies for protein conformational ensemble detection

https://doi.org/10.1016/j.sbi.2021.08.007Get rights and content

Abstract

Protein function is constrained by the three-dimensional structure but is delineated by its dynamics. This framework must satisfy specificity of function along with adaptability to changing environments and evolvability under external constraints. The accessibility of the available regions of the energy landscape for a set of conditions and shifts in the populations upon their modulation have effects propagating across scales, from biomolecular interactions, to organisms, to populations. Developing the ability to detect and juggle protein conformations supplemented by a physics-based understanding has implications for not only in vivo problems but also for resistance impeding drug discovery and bionano-sensor design.

Introduction

Let's dream the dream of the dreamer who is the protein scientist: Everything to know about the protein function from sequence, including how it responds to intrinsic and extrinsic perturbations, such as mutations and changes in the environment, would be the key to solving many biological and biotechnological problems. The questions would arrive in stages of increased complexity. Our scientist would first wish to know what the protein structure would look like under some condition it is functional. Then would come the question of how the function is facilitated, and if there are alternative conformations to achieve function. How would the protein behave in a given environment? An alternative environment? How would it respond to arising mutations? To molecular crowding? What would the role of kinetics be while, for example, switching between conformations or when the environment is altered? There is a plethora of computational strategies developed to date, tackling with each of these questions.

Making the dream come true needs unification. We begin by collecting all these inquiries into a ‘dynamical conformational search problem of proteins.’ The hypothetical free energy surface of a protein (Figure 1) consists of one or more minima which correspond to the regions defined by the various three-dimensional structures assumed by the protein. There are also other low-lying minima that represent misfolded states and higher energy local minima for on-pathway partially folded states on this surface. Because the folding free energy difference is typically on the order of a few kcal/mol for a globular protein [1], an order of magnitude smaller than that for the simplest of pairwise interactions at the covalent bond, a physics-based solution to the protein folding problem has proven extremely difficult. Small deviations at the level of bonds would completely mislead the global solution for determining the correct folded state(s) of the polypeptide chain. Recent breakthroughs applying artificial intelligence to put the structure in the vicinity of the functional region from sequence information alone allows us to set that part of the problem aside [2]. Here, we will address advances and open problems in strategies to navigate the neighborhood of the functionally relevant regions of the dynamical conformational space (Figure 1b–d). We refer the reader to the toy model of Figure 2a, while we discuss the strategies for efficient modeling of the conformational surface (CS) of proteins.

Section snippets

Atomistic simulations are reliable and predictive, albeit with limited time horizon and system size

Force-fields for atomistic molecular dynamics (MD) simulations have reached that level of accuracy which allow navigating the CS (Figure 2b). For example, the kinetics of protein–protein association at atomistic detail can now be predicted at experimentally consistent precision [3] using MD with the aid of enhanced sampling techniques and Markov state models [4]. The range of application systems is also ever-growing, including soluble, membrane, intrinsically disordered proteins, and those with

Coarse-grained approaches provide physics-based insight on protein equilibrium and dynamics

One obvious extension of full atomistic MD is to carry out simulations on coarse-grained beads. To date, Martini force-field has proven to display good approximations to experimental data, especially for membrane proteins and, with its latest update Martini 3, also for soluble proteins [13]. However, for the latter, the properties reproduced, at this time, are limited to the general protein association problems and salting in/out the behavior of soluble proteins. It has also been demonstrated

Hybrid all-atom–coarse grained methodologies extend horizon for navigating the dynamical landscape

Our understanding of exploring a dynamically changing CS has been shaped by the search for allosteric positions to engineer control into the protein structures [32] and more recently by the search for cryptic sites [33]. The former relies on finding distal positions affecting function at a known active site. The latter indicates dynamically transient and functionally relevant states that remain completely inaccessible in PDB structures and might exist unnoticed in long MD trajectories. Applying

Complete navigation of the dynamical landscape is practicable by integrative computational-experimental models

Historically, experimental validation of predictions has been crucial for the computational biology community. It is now clear that by charting the CS of proteins, computational techniques will increasingly guide experimental approaches, for example, suggesting spin labeling positions [41] and optimal positions for enzyme redesign [42]. A holistic approach towards mapping the CS of a protein has already initiated applications in determining allosteric and cryptic sites by complementing

Conflict of interest statement

Nothing declared.

Acknowledgements

The authors gratefully acknowledge support from the Scientific and Technological Research Council of Turkey (Grant Number 116F229).

References (55)

  • A.N. Lupas et al.

    The breakthrough in protein structure prediction

    Biochem J

    (2021)
  • N. Plattner et al.

    Complete protein-protein association kinetics in atomic detail revealed by molecular dynamics simulations and Markov modelling

    Nat Chem

    (2017)
  • E. Suárez et al.

    What Markov state models can and cannot do: correlation versus path-based observables in protein-folding models

    J Chem Theor Comput

    (2021)
  • A. Croitoru et al.

    Additive CHARMM36 force field for nonstandard amino acids

    J Chem Theor Comput

    (2021)
  • S. Doerr et al.

    TorchMD: a deep learning framework for molecular simulations

    J Chem Theor Comput

    (2021)
  • A. Glielmo et al.

    Unsupervised learning methods for molecular simulation data

    Chem Rev

    (2021)
  • J.R. Allison

    Computational methods for exploring protein conformations

    Biochem Soc Trans

    (2020)
  • F. Hooft et al.

    Discovering collective variables of molecular transitions via genetic algorithms and neural networks

    J Chem Theor Comput

    (2021)
  • L. Raich et al.

    Discovery of a hidden transient state in all bromodomain families

    Proc Natl Acad Sci U S A

    (2021)
  • S.K. Huang et al.

    Delineating the conformational landscape of the adenosine A(2A) receptor during G protein coupling

    Cell

    (2021)
  • J. Li et al.

    Reply to: insufficient evidence for ageing in protein dynamics

    Nat Phys

    (2021)
  • P.C.T. Souza et al.

    Martini 3: a general purpose force field for coarse-grained molecular dynamics

    Nat Methods

    (2021)
  • P.C.T. Souza et al.

    Protein-ligand binding with the coarse-grained Martini model

    Nat Commun

    (2020)
  • A. Shinobu et al.

    Coarse-grained modeling of multiple pathways in conformational transitions of multi-domain proteins

    J Chem Inf Model

    (2021)
  • F. Tama et al.

    Conformational change of proteins arising from normal mode calculations

    Protein Eng

    (2001)
  • W. Zheng

    Predicting cryptic ligand binding sites based on normal modes guided conformational sampling

    Protein-Struct Funct Bioinf

    (2021)
  • A.R. Atilgan et al.

    Anisotropy of fluctuation dynamics of proteins with an elastic network model

    Biophys J

    (2001)
  • Cited by (6)

    View full text