Finding Conformational Transition Pathways from Discrete Molecular Dynamics Simulations

Representation of Protein Dynamics Disentangled by Time-Structure-Based Prior

Representation learning (RL) is a universal technique for deriving low-dimensional disentangled representations from high-dimensional observations, aiding in a multitude of downstream tasks. RL has been extensively applied to various data types, including images and natural language. Here, we analyze molecular dynamics (MD) simulation data of biomolecules in terms of RL. Currently, state-of-the-art RL techniques, mainly motivated by the variational principle, try to capture slow motions in the representation (latent) space. Here, we propose two methods based on an alternative perspective on the disentanglement in the latent space. By disentanglement, we here mean the separation of underlying factors in the simulation data, aiding in detecting physically important coordinates for conformational transitions. The proposed methods introduce a simple prior that imposes temporal constraints in the latent space, serving as a regularization term to facilitate the capture of disentangled representations of dynamics. Comparison with other methods via the analysis of MD simulation trajectories for alanine dipeptide and chignolin validates that the proposed methods construct Markov state models (MSMs) whose implied time scales are comparable to those of the state-of-the-art methods. Using a measure based on total variation, we quantitatively evaluated that the proposed methods successfully disentangle physically important coordinates, aiding the interpretation of folding/unfolding transitions of chignolin. Overall, our methods provide good representations of complex biomolecular dynamics for downstream tasks, allowing for better interpretations of the conformational transitions.

Molecular simulation unravels the amyloidogenic misfolding of nascent ApoA1 protein, driven by deleterious point mutations occurring in between 170-178 hotspot region

Protein ApoA1 is extensively studied for its role in lipid metabolism. Its seedy dark side of amyloid formulation remains relatively understudied yet. Due to genetic mutations, the protein pathologically misshapes into its amyloid form that gets accumulated in various organs, including the heart. To contrive effective therapeutics against this debilitating congenital disorder, it is imperative to comprehend the structural ramifications induced by mutations in APoA1's dynamic conformation. Till now, several point mutations have been implicated in ApoA1's amyloidosis, although only a handful has been examined considerably. Especially, the single nucleotide polymorphisms (SNPs) that occur in-between 170-178 mutation hotspot site of APoA1 needs to be investigated, since most of them are culpable of amyloid deposition in the heart. To that effect, in the present study, we have computationally quantified and studied the ApoA1's biomolecular modifications fostered by SNPs in the 170-178 mutation hotspot. Findings from discrete molecular dynamics simulation studies indicate that the SNPs have noticeably steered the ApoA1's behaviour from its native structural dynamics. Analysis of protein's secondary structural changes exhibits a considerable change upon mutations. Further, subjecting the protein structures to simulated thermal denaturation shows increased resistance to denaturation among mutants when compared to native. Further, normal mode analysis of protein's dynamic motion also shows discrepancy in its dynamic structural change upon SNP. These structural digressions induced by SNPs can very well be the biomolecular incendiary that drives ApoA1 into its amyloidogenesis. And, understanding these structural modifications initiates a better understanding of SNP's amyloidogenic pathology on APoA1.Communicated by Ramaswamy H. Sarma.

Pre-exascale HPC approaches for molecular dynamics simulations. Covid-19 research: A use case

Exascale computing has been a dream for ages and is close to becoming a reality that will impact how molecular simulations are being performed, as well as the quantity and quality of the information derived for them. We review how the biomolecular simulations field is anticipating these new architectures, making emphasis on recent work from groups in the BioExcel Center of Excellence for High Performance Computing. We exemplified the power of these simulation strategies with the work done by the HPC simulation community to fight Covid-19 pandemics. This article is categorized under:Data Science > Computer Algorithms and ProgrammingData Science > Databases and Expert SystemsMolecular and Statistical Mechanics > Molecular Dynamics and Monte-Carlo Methods.

Exploring Conformational Space with Thermal Fluctuations Obtained by Normal-Mode Analysis

Proteins in their native states can be represented as ensembles of conformers in dynamical equilibrium. Thermal fluctuations are responsible for transitions between these conformers. Normal-modes analysis (NMA) using elastic network models (ENMs) provides an efficient procedure to explore global dynamics of proteins commonly associated with conformational transitions. In the present work, we present an iterative approach to explore protein conformational spaces by introducing structural distortions according to their equilibrium dynamics at room temperature. The approach can be used either to perform unbiased explorations of conformational space or to explore guided pathways connecting two different conformations, e.g., apo and holo forms. In order to test its performance, four proteins with different magnitudes of structural distortions upon ligand binding have been tested. In all cases, the conformational selection model has been confirmed and the conformational space between apo and holo forms has been encompassed. Different strategies have been tested that impact on the efficiency either to achieve a desired conformational change or to achieve a balanced exploration of the protein conformational multiplicity.

Elevated amyloidoses of human IAPP and amyloid beta by lipopolysaccharide and their mitigation by carbon quantum dots

Type 2 diabetes (T2D) and Alzheimer's disease (AD) represent two most prevalent amyloid diseases with a significant global burden. Pathologically, T2D and AD are characterized by the presence of amyloid plaques consisting primarily of toxic human islet amyloid polypeptide (IAPP) and amyloid beta (Aβ). It has been recently revealed that the gut microbiome plays key functions in the pathological progression of neurological disorders through the production of bacterial endotoxins, such as lipopolysaccharide (LPS). In this study, we examined the catalytic effects of LPS on IAPP and Aβ amyloidoses, and further demonstrated their mitigation with zero-dimensional carbon quantum dots (CQDs). Whereas LPS displayed preferred binding with the N-terminus of IAPP and the central hydrophobic core and C-terminus of Aβ, CQDs exhibited propensities for the amyloidogenic and C-terminus regions of IAPP and the N-terminus of Aβ, accordingly. The inhibitory effect of CQDs was verified by an embryonic zebrafish model exposed to the peptides and LPS, where impaired embryonic hatching was rescued and production of reactive oxygen species in the organism was suppressed by the nanomaterial. This study revealed a robust synergy between LPS and amyloid peptides in toxicity induction, and implicated CQDs as a potential therapeutic against the pathologies of T2D and AD.

Large-Scale Conformational Changes and Protein Function: Breaking the in silico Barrier

Large-scale conformational changes are essential to link protein structures with their function at the cell and organism scale, but have been elusive both experimentally and computationally. Over the past few years developments in cryo-electron microscopy and crystallography techniques have started to reveal multiple snapshots of increasingly large and flexible systems, deemed impossible only short time ago. As structural information accumulates, theoretical methods become central to understand how different conformers interconvert to mediate biological function. Here we briefly survey current in silico methods to tackle large conformational changes, reviewing recent examples of cross-validation of experiments and computational predictions, which show how the integration of different scale simulations with biological information is already starting to break the barriers between the in silico, in vitro, and in vivo worlds, shedding new light onto complex biological problems inaccessible so far.

Significance of triple torsional correlations in proteins

The free energy landscape (FEL) of a given complex molecular system is fundamentally the joint probability density of its many comprising degrees of freedom (DOFs). Computation of a complete FEL at atomistic scale is unfortunately intractable for a typical biomolecular system. The challenge of entropy calculation comes from various correlations among different DOFs. The common strategy to treat such complexity is expansion of the full correlation into various orders of local correlations. In reality, expansion is usually cut off at the second order (i.e. pairwise interactions) for protein torsional correlations without reliable estimation of the resulting error. Here, we estimated the mutual information of different torsion sets and found that triple correlations were significant for both local/distant residue pairs and consecutive backbone torsional segments. As expected, the third order approximations were found to be consistently better than the second order approximations. These findings were true for all analyzed proteins with different folds, were independent of the two different force fields utilized to generate trajectory sets, and were therefore likely to be of general importance for proteins. Additionally, binning strategies are of universal importance for numerical computation of correlations, we here provided a detailed comparison between equal-width and equal-sample binning for different bin numbers and demonstrated the impact of binning strategies on variances and biases of calculated mutual information. Our observation suggested that caution should be taken when quantitative comparison of correlations were intended between different studies with different binning strategies.

Torsional mutual information for 10 typical residue pairs calculated with full joint distributions (MI), second order expansion (MI₂), third order expansions (MI₃), and their linear recombinations (MI_LR).

A probabilistic network model for structural transitions in biomolecules

Biological macromolecules often undergo large conformational rearrangements during a functional cycle. To simulate these structural transitions with full atomic detail typically demands extensive computational resources. Moreover, it is unclear how to incorporate, in a principled way, additional experimental information that could guide the structural transition. This article develops a probabilistic model for conformational transitions in biomolecules. The model can be viewed as a network of anharmonic springs that break, if the experimental data support the rupture of bonds. Hamiltonian Monte Carlo in internal coordinates is used to infer structural transitions from experimental data, thereby sampling large conformational transitions without distorting the structure. The model is benchmarked on a large set of conformational transitions. Moreover, we demonstrate the use of the probabilistic network model for integrative modeling of macromolecular complexes based on data from crosslinking followed by mass spectrometry.

Discrete Molecular Dynamics Approach to the Study of Disordered and Aggregating Proteins

We present a refinement of the Coarse Grained PACSAB force field for Discrete Molecular Dynamics (DMD) simulations of proteins in aqueous conditions. As the original version, the refined method provides good representation of the structure and dynamics of folded proteins but provides much better representations of a variety of unfolded proteins, including some very large, impossible to analyze by atomistic simulation methods. The PACSAB/DMD method also reproduces accurately aggregation properties, providing good pictures of the structural ensembles of proteins showing a folded core and an intrinsically disordered region. The combination of accuracy and speed makes the method presented here a good alternative for the exploration of unstructured protein systems.

Structural Determinants of Constitutive Activation of Gα Proteins: Transducin as a Paradigm

Heterotrimeric guanine nucleotide-binding proteins (Gα proteins) are intracellular nanomachines deputed to signal transduction. The switch-on process requires the release of bound GDP from a site at the interface between GTPase and helical domains. Nucleotide release is catalyzed by G protein Coupled Receptors (GPCRs). Here we investigate the functional dynamics of wild type (WT) and six constitutively active mutants (CAMs) of the Gα protein transducin (Gt) by combining atomistic molecular dynamics (MD) simulations with Maxwell-Demod discrete MD (MDdMD) simulations of the receptor-catalyzed transition between GDP-bound and nucleotide-free states. Compared to the WT, Gt CAMs increase the overall fluctuations of nucleotide and its binding site. This is accompanied by weakening of native links involving GDP, α1, the G boxes, β1-β3, and α5. Collectively, constitutive activation by the considered mutants seems to associate with weakening of the interfaces between α5 and the surrounding portions and the interface between GTPase (G) and helical (H) domains. These mutational effects associate with increases in the overall fluctuations of the G and H domains, which reflect on the collective motions of the protein. Gt CAMs, with prominence to G56P, T325A, and F332A, prioritize collective motions of the H domain overlapping with the collective motions associated with receptor-catalyzed nucleotide release. In spite of different local perturbations, the mechanisms of nucleotide exchange catalyzed by activating mutations and by receptor are expected to employ similar molecular switches in the nucleotide binding site and to share the detachment of the H domain from the G domain.

Nonlinear backbone torsional pair correlations in proteins

Protein allostery requires dynamical structural correlations. Physical origin of which, however, remain elusive despite intensive studies during last two and half decades. Based on analysis of molecular dynamics (MD) simulation trajectories for ten proteins with different sizes and folds, we found that nonlinear backbone torsional pair (BTP) correlations, which are mainly spatially long-ranged and are dominantly executed by loop residues, exist extensively in most analyzed proteins. Examination of torsional motion for correlated BTPs suggested that such nonlinear correlations are mainly associated aharmonic torsional state transitions and in some cases strongly anisotropic local torsional motion of participating torsions, and occur on widely different and relatively longer time scales. In contrast, correlations between backbone torsions in stable α helices and β strands are mainly linear and spatially short-ranged, and are more likely to associate with harmonic local torsional motion. Further analysis revealed that the direct cause of nonlinear contributions are heterogeneous linear correlations. These findings implicate a general search strategy for novel allosteric modulation sites of protein activities.

Prediction and validation of protein intermediate states from structurally rich ensembles and coarse-grained simulations

Protein conformational changes are at the heart of cell functions, from signalling to ion transport. However, the transient nature of the intermediates along transition pathways hampers their experimental detection, making the underlying mechanisms elusive. Here we retrieve dynamic information on the actual transition routes from principal component analysis (PCA) of structurally-rich ensembles and, in combination with coarse-grained simulations, explore the conformational landscapes of five well-studied proteins. Modelling them as elastic networks in a hybrid elastic-network Brownian dynamics simulation (eBDIMS), we generate trajectories connecting stable end-states that spontaneously sample the crystallographic motions, predicting the structures of known intermediates along the paths. We also show that the explored non-linear routes can delimit the lowest energy passages between end-states sampled by atomistic molecular dynamics. The integrative methodology presented here provides a powerful framework to extract and expand dynamic pathway information from the Protein Data Bank, as well as to validate sampling methods in general.

Protein conformational changes are key to a wide range of cellular functions but remain difficult to access experimentally. Here the authors describe eBDIMS, a novel approach to predict intermediates observed in structural transition pathways from experimental ensembles.

Dancing through Life: Molecular Dynamics Simulations and Network-Centric Modeling of Allosteric Mechanisms in Hsp70 and Hsp110 Chaperone Proteins

Hsp70 and Hsp110 chaperones play an important role in regulating cellular processes that involve protein folding and stabilization, which are essential for the integrity of signaling networks. Although many aspects of allosteric regulatory mechanisms in Hsp70 and Hsp110 chaperones have been extensively studied and significantly advanced in recent experimental studies, the atomistic picture of signal propagation and energetics of dynamics-based communication still remain unresolved. In this work, we have combined molecular dynamics simulations and protein stability analysis of the chaperone structures with the network modeling of residue interaction networks to characterize molecular determinants of allosteric mechanisms. We have shown that allosteric mechanisms of Hsp70 and Hsp110 chaperones may be primarily determined by nucleotide-induced redistribution of local conformational ensembles in the inter-domain regions and the substrate binding domain. Conformational dynamics and energetics of the peptide substrate binding with the Hsp70 structures has been analyzed using free energy calculations, revealing allosteric hotspots that control negative cooperativity between regulatory sites. The results have indicated that cooperative interactions may promote a population-shift mechanism in Hsp70, in which functional residues are organized in a broad and robust allosteric network that can link the nucleotide-binding site and the substrate-binding regions. A smaller allosteric network in Hsp110 structures may elicit an entropy-driven allostery that occurs in the absence of global structural changes. We have found that global mediating residues with high network centrality may be organized in stable local communities that are indispensable for structural stability and efficient allosteric communications. The network-centric analysis of allosteric interactions has also established that centrality of functional residues could correlate with their sensitivity to mutations across diverse chaperone functions. This study reconciles a wide spectrum of structural and functional experiments by demonstrating how integration of molecular simulations and network-centric modeling may explain thermodynamic and mechanistic aspects of allosteric regulation in chaperones.

Molecular dynamics simulations: advances and applications

Molecular dynamics simulations have evolved into a mature technique that can be used effectively to understand macromolecular structure-to-function relationships. Present simulation times are close to biologically relevant ones. Information gathered about the dynamic properties of macromolecules is rich enough to shift the usual paradigm of structural bioinformatics from studying single structures to analyze conformational ensembles. Here, we describe the foundations of molecular dynamics and the improvements made in the direction of getting such ensemble. Specific application of the technique to three main issues (allosteric regulation, docking, and structure refinement) is discussed.

Path Similarity Analysis: A Method for Quantifying Macromolecular Pathways

Diverse classes of proteins function through large-scale conformational changes and various sophisticated computational algorithms have been proposed to enhance sampling of these macromolecular transition paths. Because such paths are curves in a high-dimensional space, it has been difficult to quantitatively compare multiple paths, a necessary prerequisite to, for instance, assess the quality of different algorithms. We introduce a method named Path Similarity Analysis (PSA) that enables us to quantify the similarity between two arbitrary paths and extract the atomic-scale determinants responsible for their differences. PSA utilizes the full information available in 3N-dimensional configuration space trajectories by employing the Hausdorff or Fréchet metrics (adopted from computational geometry) to quantify the degree of similarity between piecewise-linear curves. It thus completely avoids relying on projections into low dimensional spaces, as used in traditional approaches. To elucidate the principles of PSA, we quantified the effect of path roughness induced by thermal fluctuations using a toy model system. Using, as an example, the closed-to-open transitions of the enzyme adenylate kinase (AdK) in its substrate-free form, we compared a range of protein transition path-generating algorithms. Molecular dynamics-based dynamic importance sampling (DIMS) MD and targeted MD (TMD) and the purely geometric FRODA (Framework Rigidity Optimized Dynamics Algorithm) were tested along with seven other methods publicly available on servers, including several based on the popular elastic network model (ENM). PSA with clustering revealed that paths produced by a given method are more similar to each other than to those from another method and, for instance, that the ENM-based methods produced relatively similar paths. PSA was applied to ensembles of DIMS MD and FRODA trajectories of the conformational transition of diphtheria toxin, a particularly challenging example. For the AdK transition, the new concept of a Hausdorff-pair map enabled us to extract the molecular structural determinants responsible for differences in pathways, namely a set of conserved salt bridges whose charge-charge interactions are fully modelled in DIMS MD but not in FRODA. PSA has the potential to enhance our understanding of transition path sampling methods, validate them, and to provide a new approach to analyzing conformational transitions.

Many proteins are nanomachines that perform mechanical or chemical work by changing their three-dimensional shape and cycle between multiple conformational states. Computer simulations of such conformational transitions provide mechanistic insights into protein function but such simulations have been challenging. In particular, it is not clear how to quantitatively compare current simulation methods or to assess their accuracy. To that end, we present a general and flexible computational framework for quantifying transition paths—by measuring mutual geometric similarity—that, compared with existing approaches, requires minimal a-priori assumptions and can take advantage of full atomic detail alongside heuristic information derived from intuition. Using our Path Similarity Analysis (PSA) framework in parallel with several existing quantitative approaches, we examine transitions generated for a toy model of a transition and two biological systems, the enzyme adenylate kinase and diphtheria toxin. Our results show that PSA enables the quantitative comparison of different path sampling methods and aids the identification of potentially important atomistic motions by exploiting geometric information in transition paths. The method has the potential to enhance our understanding of transition path sampling methods, validate them, and to provide a new approach to analyzing macromolecular conformational transitions.

Correlated motions are a fundamental property of β-sheets

Correlated motions in proteins can mediate fundamental biochemical processes such as signal transduction and allostery. The mechanisms that underlie these processes remain largely unknown due mainly to limitations in their direct detection. Here, based on a detailed analysis of protein structures deposited in the protein data bank, as well as on state-of-the art molecular simulations, we provide general evidence for the transfer of structural information by correlated backbone motions, mediated by hydrogen bonds, across β-sheets. We also show that the observed local and long-range correlated motions are mediated by the collective motions of β-sheets and investigate their role in large-scale conformational changes. Correlated motions represent a fundamental property of β-sheets that contributes to protein function.

A theoretical view of protein dynamics

Proteins are fascinating supramolecular structures, which are able to recognize ligands transforming binding information into chemical signals. They can transfer information across the cell, can catalyse complex chemical reactions, and are able to transform energy into work with much more efficiency than any human engine. The unique abilities of proteins are tightly coupled with their dynamic properties, which are coded in a complex way in the sequence and carefully refined by evolution. Despite its importance, our experimental knowledge of protein dynamics is still rather limited, and mostly derived from theoretical calculations. I will review here, in a systematic way, the current state-of-the-art theoretical approaches to the study of protein dynamics, emphasizing the most recent advances, examples of use and the expected lines of development in the near future.

Exploring the conformational transitions of biomolecular systems using a simple two-state anisotropic network model

Biomolecular conformational transitions are essential to biological functions. Most experimental methods report on the long-lived functional states of biomolecules, but information about the transition pathways between these stable states is generally scarce. Such transitions involve short-lived conformational states that are difficult to detect experimentally. For this reason, computational methods are needed to produce plausible hypothetical transition pathways that can then be probed experimentally. Here we propose a simple and computationally efficient method, called ANMPathway, for constructing a physically reasonable pathway between two endpoints of a conformational transition. We adopt a coarse-grained representation of the protein and construct a two-state potential by combining two elastic network models (ENMs) representative of the experimental structures resolved for the endpoints. The two-state potential has a cusp hypersurface in the configuration space where the energies from both the ENMs are equal. We first search for the minimum energy structure on the cusp hypersurface and then treat it as the transition state. The continuous pathway is subsequently constructed by following the steepest descent energy minimization trajectories starting from the transition state on each side of the cusp hypersurface. Application to several systems of broad biological interest such as adenylate kinase, ATP-driven calcium pump SERCA, leucine transporter and glutamate transporter shows that ANMPathway yields results in good agreement with those from other similar methods and with data obtained from all-atom molecular dynamics simulations, in support of the utility of this simple and efficient approach. Notably the method provides experimentally testable predictions, including the formation of non-native contacts during the transition which we were able to detect in two of the systems we studied. An open-access web server has been created to deliver ANMPathway results.

Many biomolecules are like tiny molecular machines that need to change their shapes and visit many states to perform their biological functions. For a complete molecular understanding of a biological process, one needs to have information on the relevant stable states of the system in question, as well as the pathways by which the system travels from one state to another. We report here an efficient computational method that uses the knowledge of experimental structures of a pair of stable states in order to construct an energetically favoravle pathway between them. We adopt a simple representation of the molecular system by replacing the atoms with beads connected by springs and constructing an energy function with two minima around the end-states. We searched for the structure with highest energy that the system is most likely to visit during the transition and created two paths starting from this structure and proceeding toward the end-states. The combined result of these two paths is the minimum energy pathway between the two stable states. We apply this method to study important structural changes in one enzyme and three large proteins that transport small molecules and ions across the cell membrane.

Exploration of conformational transition pathways from coarse-grained simulations

MOTIVATION

A new algorithm to trace conformational transitions in proteins is presented. The method uses discrete molecular dynamics as engine to sample protein conformational space. A multiple minima Go-like potential energy function is used in combination with several enhancing sampling strategies, such as metadynamics, Maxwell Demon molecular dynamics and essential dynamics. The method, which shows an unprecedented computational efficiency, is able to trace a wide range of known experimental transitions. Contrary to simpler methods our strategy does not introduce distortions in the chemical structure of the protein and is able to reproduce well complex non-linear conformational transitions. The method, called GOdMD, can easily introduce additional restraints to the transition (presence of ligand, known intermediate, known maintained contacts, …) and is freely distributed to the community through the Spanish National Bioinformatics Institute (http://mmb.irbbarcelona.org/GOdMD).

AVAILABILITY

Freely available on the web at http://mmb.irbbarcelona.org/GOdMD.

Efficient Relaxation of Protein-Protein Interfaces by Discrete Molecular Dynamics Simulations

Protein-protein interactions are responsible for the transfer of information inside the cell and represent one of the most interesting research fields in structural biology. Unfortunately, after decades of intense research, experimental approaches still have difficulties in providing 3D structures for the hundreds of thousands of interactions formed between the different proteins in a living organism. The use of theoretical approaches like docking aims to complement experimental efforts to represent the structure of the protein interactome. However, we cannot ignore that current methods have limitations due to problems of sampling of the protein-protein conformational space and the lack of accuracy of available force fields. Cases that are especially difficult for prediction are those in which complex formation implies a non-negligible change in the conformation of the interacting proteins, i.e., those cases where protein flexibility plays a key role in protein-protein docking. In this work, we present a new approach to treat flexibility in docking by global structural relaxation based on ultrafast discrete molecular dynamics. On a standard benchmark of protein complexes, the method provides a general improvement over the results obtained by rigid docking. The method is especially efficient in cases with large conformational changes upon binding, in which structure relaxation with discrete molecular dynamics leads to a predictive success rate double that obtained with state-of-the-art rigid-body docking.