Dynamic Ensembles Research Articles

Heterogeneous Slowdown of Dynamics in the Condensate of an Intrinsically Disordered Protein.

The high concentration of proteins and other biological macromolecules inside biomolecular condensates leads to dense and confined environments, which can affect the dynamic ensembles and the time scales of the conformational transitions. Here, we use atomistic molecular dynamics (MD) simulations of the intrinsically disordered low complexity domain (LCD) of the human fused in sarcoma (FUS) RNA-binding protein to study how self-crowding inside a condensate affects the dynamic motions of the protein. We found a heterogeneous retardation of the protein dynamics in the condensate with respect to the dilute phase, with large-amplitude motions being strongly slowed by up to 2 orders of magnitude, whereas small-scale motions, such as local backbone fluctuations and side-chain rotations, are less affected. The results support the notion of a liquid-like character of the condensates and show that different protein motions respond differently to the environment.

SeqDance: A Protein Language Model for Representing Protein Dynamic Properties.

Proteins perform their functions by folding amino acid sequences into dynamic structural ensembles. Despite the important role of protein dynamics, their complexity and the absence of efficient representation methods have limited their integration into studies on protein function and mutation fitness, especially in deep learning applications. To address this, we present SeqDance, a protein language model designed to learn representation of protein dynamic properties directly from sequence alone. SeqDance is pre-trained on dynamic biophysical properties derived from over 30,400 molecular dynamics trajectories and 28,600 normal mode analyses. Our results show that SeqDance effectively captures local dynamic interactions, co-movement patterns, and global conformational features, even for proteins lacking homologs in the pre-training set. Additionally, we showed that SeqDance enhances the prediction of protein fitness landscapes, disorder-to-order transition binding regions, and phase-separating proteins. By learning dynamic properties from sequence, SeqDance complements conventional evolution- and static structure-based methods, offering new insights into protein behavior and function.

Integrating ytopt and libEnsemble to autotune OpenMC

Ytopt is a Python machine-learning-based autotuning software package developed within the ECP PROTEAS-TUNE project. The ytopt software adopts an asynchronous search framework that consists of sampling a small number of input parameter configurations and progressively fitting a surrogate model over the input-output space until exhausting the user-defined maximum number of evaluations or the wall-clock time. libEnsemble is a Python toolkit for coordinating workflows of asynchronous and dynamic ensembles of calculations across massively parallel resources developed within the ECP PETSc/TAO project. libEnsemble helps users take advantage of massively parallel resources to solve design, decision, and inference problems and expands the class of problems that can benefit from increased parallelism. In this paper we present our methodology and framework to integrate ytopt and libEnsemble to take advantage of massively parallel resources to accelerate the autotuning process. Specifically, we focus on using the proposed framework to autotune the ECP ExaSMR application OpenMC, an open source Monte Carlo particle transport code. OpenMC has seven tunable parameters some of which have large ranges such as the number of particles in-flight, which is in the range of 100,000 to 8 million, with its default setting of 1 million. Setting the proper combination of these parameter values to achieve the best performance is extremely time-consuming. Therefore, we apply the proposed framework to autotune the MPI/OpenMP offload version of OpenMC based on a user-defined metric such as the figure of merit (FoM) (particles/s) or energy efficiency energy-delay product (EDP) on Crusher at Oak Ridge Leadership Computing Facility. The experimental results show that we achieve the improvement up to 29.49% in FoM and up to 30.44% in EDP.

Dynamic stacking ensemble for cross-language code smell detection.

Code smells refer to poor design and implementation choices by software engineers that might affect the overall software quality. Code smells detection using machine learning models has become a popular area to build effective models that are capable of detecting different code smells in multiple programming languages. However, the process of building of such effective models has not reached a state of stability, and most of the existing research focuses on Java code smells detection. The main objective of this article is to propose dynamic ensembles using two strategies, namely greedy search and backward elimination, which are capable of accurately detecting code smells in two programming languages (i.e., Java and Python), and which are less complex than full stacking ensembles. The detection performance of dynamic ensembles were investigated within the context of four Java and two Python code smells. The greedy search and backward elimination strategies yielded different base models lists to build dynamic ensembles. In comparison to full stacking ensembles, dynamic ensembles yielded less complex models when they were used to detect most of the investigated Java and Python code smells, with the backward elimination strategy resulting in less complex models. Dynamic ensembles were able to perform comparably against full stacking ensembles with no significant detection loss. This article concludes that dynamic stacking ensembles were able to facilitate the effective and stable detection performance of Java and Python code smells over all base models and with less complexity than full stacking ensembles.

Adjusting the Energy Profile for CH-O Interactions Leads to Improved Stability of RNA Stem-Loop Structures in MD Simulations.

The role of ribonucleic acid (RNA) in biology continues to grow, but insight into important aspects of RNA behavior is lacking, such as dynamic structural ensembles in different environments, how flexibility is coupled to function, and how function might be modulated by small molecule binding. In the case of proteins, much progress in these areas has been made by complementing experiments with atomistic simulations, but RNA simulation methods and force fields are less mature. It remains challenging to generate stable RNA simulations, even for small systems where well-defined, thermostable structures have been established by experiments. Many different aspects of RNA energetics have been adjusted in force fields, seeking improvements that are transferable across a variety of RNA structural motifs. In this work, the role of weak CH···O interactions is explored, which are ubiquitous in RNA structure but have received less attention in RNA force field development. By comparing data extracted from high-resolution RNA crystal structures to energy profiles from quantum mechanics and force field calculations, it is shown that CH···O interactions are overly repulsive in the widely used Amber RNA force fields. A simple, targeted adjustment of CH···O repulsion that leaves the remainder of the force field unchanged was developed. Then, the standard and modified force fields were tested using molecular dynamics (MD) simulations with explicit water and salt, amassing over 300 μs of data for multiple RNA systems containing important features such as the presence of loops, base stacking interactions as well as canonical and noncanonical base pairing. In this work and others, standard force fields lead to reproducible unfolding of the NMR-based structures. Including a targeted CH···O adjustment in an otherwise identical protocol dramatically improves the outcome, leading to stable simulations for all RNA systems tested.

Excited-Ground-State Transition of the RNA Strand Slippage Mechanism Captured by the Base-Specific Force Field.

Excited-ground-state transition and strand slippage of RNA play key roles in transcription and translation of central dogma. Due to limitation of current experimental techniques, the dynamic structure ensembles of RNA remain inadequately understood. Molecular dynamics simulations offer a promising complementary approach, whose accuracy depends on the force field. Here, we develop the new version of RNA base-specific force field (BSFF2) to address underestimation of base pairing stability and artificial backbone conformations. Extensive evaluations on typical RNA systems have comprehensively confirmed the accuracy of BSFF2. Furthermore, BSFF2 demonstrates exceptional efficiency in de novo folding of tetraloops and reproducing base pair reshuffling transition between RNA excited and ground states. Then, we explored the RNA strand slippage mechanism with BSFF2. We conducted a comprehensive three-dimensional structural investigation into the strand slippage of the most complex r(G4C2)9 repeat element and presented the molecular details in the dynamic transition along with the underlying mechanism. Our results of capturing the strand slippage, excited-ground transition, de novo folding, and simulations for various typical RNA motifs indicate that BSFF2 should be one of valuable tools for dynamic conformation research and structure prediction of RNA, and a future contribution to RNA-targeted drug design as well as RNA therapy development.

Coupling the role of lipids to the conformational dynamics of the ABC transporter P-glycoprotein

The ATP-binding cassette transporter P-glycoprotein (P-gp) is a multidrug efflux pump that is overexpressed in a variety of cancers and associated with the drug-resistance phenomenon. P-gp structures were previously determined in detergent and in nanodiscs, in which different transmembrane helix conformations were found, “straight” and “kinked,” respectively, indicating a possible role of the lipid environment on the P-gp structural ensemble. Here, we investigate the dynamic conformational ensembles and protein-lipid interactions of two human P-gp inward-open conformers, straight and kinked, employing all-atom molecular dynamics (MD) simulations in asymmetric multicomponent lipid bilayers that mimic the highly specialized hepatocyte membrane in which P-gp is expressed. The two conformers are found to differ in terms of the accessibility of the substrate cavity. The MD simulations show how cholesterol and different lipid species wedge, snorkel, and partially enter into the cavity of the straight P-gp conformer solved in detergent. However, access to the cavity of the kinked P-gp conformer solved in nanodiscs is restricted. Furthermore, the volume and dynamic fluctuations of the substrate cavity largely differ between the two P-gp conformers and are modulated by the presence (or absence) of cholesterol in the membrane and/or of ATP. From the mechanistic perspective, the findings indicate that the straight conformer likely precedes the kinked conformer in the functional working cycle of P-gp, with the latter conformation representing a post substrate-bound state. The inaccessibility of the main transmembrane cavity in the kinked conformer might be crucial in preventing substrate disengagement and transport withdrawal. Remarkably, in our unbiased MD simulations, one transmembrane helix (TM10) of the straight conformer underwent a spontaneous transition to a kinked conformation, underlining the relevance of both conformations in a native phospholipid environment and revealing structural descriptors defining the transition between the two P-gp conformers.

FRET-guided modeling of nucleic acids.

The functional diversity of RNAs is encoded in their innate conformational heterogeneity. The combination of single-molecule spectroscopy and computational modeling offers new attractive opportunities to map structural transitions within nucleic acid ensembles. Here, we describe a framework to harmonize single-molecule Förster resonance energy transfer (FRET) measurements with molecular dynamics simulations and de novo structure prediction. Using either all-atom or implicit fluorophore modeling, we recreate FRET experiments in silico, visualize the underlying structural dynamics and quantify the reaction coordinates. Using multiple accessible-contact volumes as a post hoc scoring method for fragment assembly in Rosetta, we demonstrate that FRET can be used to filter a de novo RNA structure prediction ensemble by refuting models that are not compatible with in vitro FRET measurement. We benchmark our FRET-assisted modeling approach on double-labeled DNA strands and validate it against an intrinsically dynamic manganese(II)-binding riboswitch. We show that a FRET coordinate describing the assembly of a four-way junction allows our pipeline to recapitulate the global fold of the riboswitch displayed by the crystal structure. We conclude that computational fluorescence spectroscopy facilitates the interpretability of dynamic structural ensembles and improves the mechanistic understanding of nucleic acid interactions.

Dynamic Classification Ensembles for Handling Imbalanced Multiclass Drifted Data Streams

Machine learning models often encounter significant difficulties when dealing with multiclass imbalanced data streams in nonstationary environments. These challenges can lead to biased and unreliable predictions, which ultimately impact the overall performance of the models. To address these issues, we propose an innovative approach that integrates dynamic ensemble selection, an adaptive technique for managing imbalanced multiclass data streams, with a concept drift detector for recognizing stream changes and the K-nearest neighbor (KNN) algorithm to tackle issues related to class overlap. The primary objective was to improve the classification of imbalanced multiclass drifted data streams. The adaptive oversampling method generates synthetic samples to mitigate the issues associated with imbalanced data streams. This method utilizes KNN to ensure that the generated samples do not overlap. To handle incoming data streams, a drift detector assists in deciding whether to retain the existing classifiers or create a new one. Dynamic Ensemble Selection (DES) was utilized to select the most appropriate classifier for incoming data, aiming to optimize the performance of the classification task. The proposed method offers an effective solution for achieving an accurate and resilient classification in the context of imbalanced multiclass drifted data streams. To evaluate the effectiveness of our proposal, we conducted experiments on a variety of datasets, including benchmark datasets, real application stream datasets, and synthetic data streams. The experimental results demonstrate the superiority of our contribution in addressing the challenges posed by imbalanced multiclass drifted data streams.

What makes memories last—dynamic ensembles or static synapses?

What makes memories last—dynamic ensembles or static synapses?

Friends and relatives: insight into conformational regulation from orthologues and evolutionary lineages using KIF and KIN.

Noncovalent interaction networks provide a powerful means to represent and analyze protein structure. Such networks can represent both static structures and dynamic conformational ensembles. We have recently developed two tools for analyzing such interaction networks and generating hypotheses for protein engineering. Here, we apply these tools to the conformational regulation of substrate specificity in class A β-lactamases, particularly the evolutionary development from generalist to specialist catalytic function and how that can be recapitulated or reversed by protein engineering. These tools, KIF and KIN, generate a set of prioritized residues and interactions as targets for experimental protein engineering.

Dynamic Decision Tree Ensembles for Energy-Efficient Inference on IoT Edge Nodes

With the increasing popularity of Internet of Things (IoT) devices, there is a growing need for energy-efficient Machine Learning (ML) models that can run on constrained edge nodes. Decision tree ensembles, such as Random Forests (RFs) and Gradient Boosting (GBTs), are particularly suited for this task, given their relatively low complexity compared to other alternatives. However, their inference time and energy costs are still significant for edge hardware. Given that said costs grow linearly with the ensemble size, this paper proposes the use of dynamic ensembles, that adjust the number of executed trees based both on a latency/energy target and on the complexity of the processed input, to trade-off computational cost and accuracy. We focus on deploying these algorithms on multi-core low-power IoT devices, designing a tool that automatically converts a Python ensemble into optimized C code, and exploring several optimizations that account for the available parallelism and memory hierarchy. We extensively benchmark both static and dynamic RFs and GBTs on three state-of-the-art IoT-relevant datasets, using an 8-core ultra-lowpower System-on-Chip (SoC), GAP8, as the target platform. Thanks to the proposed early-stopping mechanisms, we achieve an energy reduction of up to 37.9% with respect to static GBTs (8.82 uJ vs 14.20 uJ per inference) and 41.7% with respect to static RFs (2.86 uJ vs 4.90 uJ per inference), without losing accuracy compared to the static model.

Representing uncertainty in limited-area data assimilating ocean models

A limited-area ocean prediction system is developed to acquire forecast error covariances related to uncertainty in atmospheric forcing and turbulent mixing using perturbed model parameters within an Ensemble Kalman Filter (EnKF). The system performs sequential data assimilation delivering realistic ocean state estimation and forecasts. It is initialised to observations using the EnKF, the hybrid-EnKF and Ensemble Optimal Interpolation (EnOI). It has higher resolution than the parent global model, includes tides and assimilates altimetric sea level anomaly to constrain the offshore mesoscale circulation. Dynamic ensemble spread introduced by parameter uncertainty shows agreement with error estimates obtained from forecast innovation statistics. Hybrid EnKF offers improvements to both EnKF and EnOI, having smaller analysis increments and forecast errors. The ensemble mean of the hybrid-EnKF is more accurate than any member due to the introduction of parameterised error which behaves as additive inflation in the hybrid EnKF system. This indicates the parameterisation acts as a non-linear filter for model forecast error growth. To demonstrate consistency, experiments are carried out for various regions that differ in their oceanographic situations. The hybrid EnKF data assimilation system enables practical use of dynamic ensembles that capture the flow dependent errors which are mixed with static or climatological covariances for situations where model error is systematically under-estimated.

Towards the transferable audio adversarial attack via ensemble methods

In recent years, deep learning (DL) models have achieved significant progress in many domains, such as autonomous driving, facial recognition, and speech recognition. However, the vulnerability of deep learning models to adversarial attacks has raised serious concerns in the community because of their insufficient robustness and generalization. Also, transferable attacks have become a prominent method for black-box attacks. In this work, we explore the potential factors that impact adversarial examples (AEs) transferability in DL-based speech recognition. We also discuss the vulnerability of different DL systems and the irregular nature of decision boundaries. Our results show a remarkable difference in the transferability of AEs between speech and images, with the data relevance being low in images but opposite in speech recognition. Motivated by dropout-based ensemble approaches, we propose random gradient ensembles and dynamic gradient-weighted ensembles, and we evaluate the impact of ensembles on the transferability of AEs. The results show that the AEs created by both approaches are valid for transfer to the black box API.

Combined NMR and molecular dynamics conformational filter identifies unambiguously dynamic ensembles of Dengue protease NS2B/NS3pro

The dengue protease NS2B/NS3pro has been reported to adopt either an ‘open’ or a ‘closed’ conformation. We have developed a conformational filter that combines NMR with MD simulations to identify conformational ensembles that dominate in solution. Experimental values derived from relaxation parameters for the backbone and methyl side chains were compared with the corresponding back-calculated relaxation parameters of different conformational ensembles obtained from free MD simulations. Our results demonstrate a high prevalence for the ‘closed’ conformational ensemble while the ‘open’ conformation is absent, indicating that the latter conformation is most probably due to crystal contacts. Conversely, conformational ensembles in which the positioning of the co-factor NS2B results in a ‘partially’ open conformation, previously described in both MD simulations and X-ray studies, were identified by our conformational filter. Altogether, we believe that our approach allows for unambiguous identification of true conformational ensembles, an essential step for reliable drug discovery.

Lactadherin's Multistate Binding Predicts Stable Membrane-Bound Conformations of Factors V and VIII's C Domains.

Protein binding to negatively charged lipids is essential for maintaining numerous vital cellular processes where its dysfunction can lead to various diseases. One such protein that plays a crucial role in this process is lactadherin, which competes with coagulation factors for membrane binding sites to regulate blood clotting. Despite identifying key binding regions of these proteins through structural and biochemical studies, models incorporating membrane dynamics are still lacking. In this study, we report on the multimodal binding of lactadherin and use it to gain insight into the binding mechanisms of its C domain homologs, factor V and factor VIII. Molecular dynamics simulations enhanced with the highly mobile mimetic model enabled the determination of lactadherin's multimodal binding on membranes that revealed critical interacting residues consistent with prior NMR and mutagenesis data. The binding occurred primarily via two dynamic structural ensembles: an inserted state and an unreported, highly conserved side-lying state driven by a cationic patch. We utilized these findings to analyze the membrane binding domains of coagulation factors V and VIII and identified their preferred membrane-bound conformations. Specifically, factor V's C domains maintained an inserted state, while factor VIII preferred a tilted, side-lying state that permitted antibody binding. Insight into lactadherin's atomistically resolved membrane interactions from a multistate perspective can guide new therapeutic opportunities in treating diseases related to blood coagulation.

Mixed, nonclassical behavior in a classic allosteric protein

Allostery is a major driver of biological processes requiring coordination. Thus, it is one of the most fundamental and remarkable phenomena in nature, and there is motivation to understand and manipulate it to a multitude of ends. Today, it is often described in terms of two phenomenological models proposed more than a half-century ago involving only T(tense) or R(relaxed) conformations. Here, methyl-based NMR provides extensive detail on a dynamic T to R switch in the classical dimeric allosteric protein, yeast chorismate mutase (CM), that occurs in the absence of substrate, but only with the activator bound. Switching of individual subunits is uncoupled based on direct observation of mixed TR states in the dimer. This unique finding excludes both classic models and solves the paradox of a coexisting hyperbolic binding curve and highly skewed substrate-free T-R equilibrium. Surprisingly, structures of the activator-bound and effector-free forms of CM appear the same by NMR, providing another example of the need to account for dynamic ensembles. The apo enzyme, which has a sigmoidal activity profile, is shown to switch, not to R, but to a related high-energy state. Thus, the conformational repertoire of CM does not just change as a matter of degree depending on the allosteric input, be it effector and/or substrate. Rather, the allosteric model appears to completely change in different contexts, which is only consistent with modern ensemble-based frameworks.

Probing the substrate binding modes and catalytic mechanisms of BLEG-1, a promiscuous B3 metallo-β-lactamase with glyoxalase II properties.

BLEG-1 from Bacillus lehensis G1 is an evolutionary divergent B3 metallo-β-lactamase (MBL) that exhibited both β-lactamase and glyoxalase II (GLXII) activities. Sequence, phylogeny, biochemical and structural relatedness of BLEG-1 to B3 MBL and GLXII suggested BLEG-1 might be an intermediate in the evolutionary path of B3 MBL from GLXII. The unique active site cavity of BLEG-1 that recognizes both β-lactam antibiotics and S-D-lactoylglutathione (SLG) had been postulated as the key factor for its dual activity. In this study, dynamic ensembles of BLEG-1 and its substrate complexes divulged conformational plasticity and binding modes of structurally distinct substrates to the enzyme, providing better insights into its structure-to-function relationship and enzymatic promiscuity. Our results highlight the flexible nature of the active site pocket of BLEG-1, which is governed by concerted loop motions involving loop7+α3+loop8 and loop12 around the catalytic core, thereby moulding the binding pocket and facilitate interactions of BLEG-1 with both ampicillin and SLG. The distribution of (i) predominantly hydrophobic amino acids in the N-terminal domain, and (ii) flexible amino acids with polar and/or charged side chains in both N- and C-termini provide additional advantages to BLEG-1 in confining the aromatic group of ampicillin, and polar groups of SLG, respectively. The importance of these residues for substrates binding was further confirmed by the reduction in MBL and GLXII activities upon alanine substitutions of Ile-10, Phe-57, Arg-94, Leu-95, and Arg-159. Based on molecular dynamics simulation, mutational, and biochemical data presented herein, the catalytic mechanisms of BLEG-1 toward the hydrolysis of β-lactams and SLG were proposed.

Cross-Linked α-Synuclein as Inhibitor of Amyloid Formation.

The aggregation and amyloid formation of α-synuclein is associated with Parkinson's disease and other synucleinopathies. In its native, monomeric form α-synuclein is an intrinsically disordered protein represented by highly dynamic conformational ensembles. Inhibition of α-synuclein aggregation using small molecules, peptides, or proteins has been at the center of interest in recent years. Our aim was to explore the effects of cross-linking on the structure and aggregation/amyloid formation properties of α-synuclein. Comparative analysis of available high-resolution amyloid structures and representative structural models and MD trajectory of monomeric α-synuclein revealed that potential cross-links in the monomeric protein are mostly incompatible with the amyloid forms and thus might inhibit fibrillation. Monomeric α-synuclein has been intramolecularly chemically cross-linked under various conditions using different cross-linkers. We determined the location of cross-links and their frequency using mass spectrometry and found that most of them cannot be realized in the amyloid structures. The inhibitory potential of cross-linked proteins has been experimentally investigated using various methods, including thioflavin-T fluorescence and transmission electron microscopy. We found that conformational constraints applied by cross-linking fully blocked α-synuclein amyloid formation. Moreover, DTSSP-cross-linked molecules exhibited an inhibitory effect on the aggregation of unmodified α-synuclein as well.

Detecting dynamic domains and local fluctuations in complex molecular systems via timelapse neighbors shuffling

It is known that the behavior of many complex systems is controlled by local dynamic rearrangements or fluctuations occurring within them. Complex molecular systems, composed of many molecules interacting with each other in a Brownian storm, make no exception. Despite the rise of machine learning and of sophisticated structural descriptors, detecting local fluctuations and collective transitions in complex dynamic ensembles remains often difficult. Here, we show a machine learning framework based on a descriptor which we name Local Environments and Neighbors Shuffling (LENS), that allows identifying dynamic domains and detecting local fluctuations in a variety of systems in an abstract and efficient way. By tracking how much the microscopic surrounding of each molecular unit changes over time in terms of neighbor individuals, LENS allows characterizing the global (macroscopic) dynamics of molecular systems in phase transition, phases-coexistence, as well as intrinsically characterized by local fluctuations (e.g., defects). Statistical analysis of the LENS time series data extracted from molecular dynamics trajectories of, for example, liquid-like, solid-like, or dynamically diverse complex molecular systems allows tracking in an efficient way the presence of different dynamic domains and of local fluctuations emerging within them. The approach is found robust, versatile, and applicable independently of the features of the system and simply provided that a trajectory containing information on the relative motion of the interacting units is available. We envisage that "such a LENS" will constitute a precious basis for exploring the dynamic complexity of a variety of systems and, given its abstract definition, not necessarily of molecular ones.