Ensemble Of Proteins Research Articles

Recent advances in protein conformation sampling by combining machine learning with molecular simulation

Abstract The rapid advancement and broad application of machine learning (ML) have driven a groundbreaking revolution in computational biology. One of the most cutting-edge and important applications of ML is its integration with molecular simulations to improving the sampling efficiency of the vast conformational space of large biomolecules. This review focuses on recent studies that utilize ML-based techniques in the exploration of protein conformational landscape. We first highlight the recent development of ML-aided enhanced sampling methods, including heuristic algorithms and neural networks that are designed to refine the selection of reaction coordinates for the construction of bias potential, or facilitate the exploration of the unsampled region of the energy landscape. Further, we review the development of autoencoder based methods which combine molecular simulations and deep learning to expand the search of protein conformations. Lastly, we discuss the cutting-edge methodologies for the one-shot generation of protein conformations with precise Boltzmann weights. Collectively, this review demonstrates the promising potential of machine learning in revolutionizing our insight into the complex conformational ensembles of proteins.

Structure and computation-guided yeast surface display for the evolution of TIMP-based matrix metalloproteinase inhibitors.

The study of protein-protein interactions (PPIs) and the engineering of protein-based inhibitors often employ two distinct strategies. One approach leverages the power of combinatorial libraries, displaying large ensembles of mutant proteins, for example, on the yeast cell surface, to select binders. Another approach harnesses computational modeling, sifting through an astronomically large number of protein sequences and attempting to predict the impact of mutations on PPI binding energy. Individually, each approach has inherent limitations, but when combined, they generate superior outcomes across diverse protein engineering endeavors. This synergistic integration of approaches aids in identifying novel binders and inhibitors, fine-tuning specificity and affinity for known binding partners, and detailed mapping of binding epitopes. It can also provide insight into the specificity profiles of varied PPIs. Here, we outline strategies for directing the evolution of tissue inhibitors of metalloproteinases (TIMPs), which act as natural inhibitors of matrix metalloproteinases (MMPs). We highlight examples wherein design of combinatorial TIMP libraries using structural and computational insights and screening these libraries of variants using yeast surface display (YSD), has successfully optimized for MMP binding and selectivity, and conferred insight into the PPIs involved.

DiAMoNDBack: Diffusion-Denoising Autoregressive Model for Non-Deterministic Backmapping of Cα Protein Traces.

Coarse-grained molecular models of proteins permit access to length and time scales unattainable by all-atom models and the simulation of processes that occur on long time scales, such as aggregation and folding. The reduced resolution realizes computational accelerations, but an atomistic representation can be vital for a complete understanding of mechanistic details. Backmapping is the process of restoring all-atom resolution to coarse-grained molecular models. In this work, we report DiAMoNDBack (Diffusion-denoising Autoregressive Model for Non-Deterministic Backmapping) as an autoregressive denoising diffusion probability model to restore all-atom details to coarse-grained protein representations retaining only Cα coordinates. The autoregressive generation process proceeds from the protein N-terminus to C-terminus in a residue-by-residue fashion conditioned on the Cα trace and previously backmapped backbone and side-chain atoms within the local neighborhood. The local and autoregressive nature of our model makes it transferable between proteins. The stochastic nature of the denoising diffusion process means that the model generates a realistic ensemble of backbone and side-chain all-atom configurations consistent with the coarse-grained Cα trace. We train DiAMoNDBack over 65k+ structures from the Protein Data Bank (PDB) and validate it in applications to a hold-out PDB test set, intrinsically disordered protein structures from the Protein Ensemble Database (PED), molecular dynamics simulations of fast-folding mini-proteins from DE Shaw Research, and coarse-grained simulation data. We achieve state-of-the-art reconstruction performance in terms of correct bond formation, avoidance of side-chain clashes, and the diversity of the generated side-chain configurational states. We make the DiAMoNDBack model publicly available as a free and open-source Python package.

PED in 2024: improving the community deposition of structural ensembles for intrinsically disordered proteins.

The Protein Ensemble Database (PED) (URL: https://proteinensemble.org) is the primary resource for depositing structural ensembles of intrinsically disordered proteins. This updated version of PED reflects advancements in the field, denoting a continual expansion with a total of 461 entries and 538 ensembles, including those generated without explicit experimental data through novel machine learning (ML) techniques. With this significant increment in the number of ensembles, a few yet-unprecedented new entries entered the database, including those also determined or refined by electron paramagnetic resonance or circular dichroism data. In addition, PED was enriched with several new features, including a novel deposition service, improved user interface, new database cross-referencing optionsand integration with the 3D-Beacons network-all representing efforts to improve the FAIRness of the database. Foreseeably, PED will keep growing in size and expanding with new types of ensembles generated by accurate and fast ML-based generative models and coarse-grained simulations. Therefore, among future efforts, priority will be given to further develop the database to be compatible with ensembles modeled at a coarse-grained level.

Accurate and Fast Prediction of Intrinsically Disordered Protein by Multiple Protein Language Models and Ensemble Learning.

Intrinsically disordered proteins (IDPs) play a vital role in various biological processes and have attracted increasing attention in the past few decades. Predicting IDPs from the primary structures of proteins offers a rapid and facile means of protein analysis without necessitating crystal structures. In particular, machine learning methods have demonstrated their potential in this field. Recently, protein language models (PLMs) are emerging as a promising approach to extracting essential information from protein sequences and have been employed in protein modeling to utilize their advantages of precision and efficiency. In this article, we developed a novel IDP prediction method named IDP-ELM to predict the intrinsically disordered regions (IDRs) as well as their functions including disordered flexible linkers and disordered protein binding. This method utilizes high-dimensional representations extracted from several state-of-the-art PLMs and predicts IDRs by ensemble learning based on bidirectional recurrent neural networks. The performance of the method was evaluated on two independent test data sets from CAID (critical assessment of protein intrinsic disorder prediction) and CAID2, indicating notable improvements in terms of area under the receiver operating characteristic (AUC), Matthew's correlation coefficient (MCC), and F1 score. Moreover, IDP-ELM requires solely protein sequences as inputs and does not entail a time-consuming process of protein profile generation, which is a prerequisite for most existing state-of-the-art methods, enabling an accurate, fast, and convenient tool for proteome-level analysis. The corresponding reproducible source code and model weights are available at https://github.com/xu-shi-jie/idp-elm.

Form factor determination of biological molecules with X-ray free electron laser small-angle scattering (XFEL-SAS)

Free-electron lasers (FEL) are revolutionizing X-ray-based structural biology methods. While protein crystallography is already routinely performed at FELs, Small Angle X-ray Scattering (SAXS) studies of biological macromolecules are not as prevalent. SAXS allows the study of the shape and overall structure of proteins and nucleic acids in solution, in a quasi-native environment. In solution, chemical and biophysical parameters that have an influence on the structure and dynamics of molecules can be varied and their effect on conformational changes can be monitored in time-resolved XFEL and SAXS experiments. We report here the collection of scattering form factors of proteins in solution using FEL X-rays. The form factors correspond to the scattering signal of the protein ensemble alone; the scattering contributions from the solvent and the instrument are separately measured and accurately subtracted. The experiment was done using a liquid jet for sample delivery. These results pave the way for time-resolved studies and measurements from dilute samples, capitalizing on the intense and short FEL X-ray pulses.

Functional Interactions of BAF and LEM Proteins in the Formation of Germ Cells

Recovery of the nuclear structure after cell division requires special interactions between the integral proteins of the inner nuclear membrane having a special LEM domain (LEMD), nuclear lamina proteins (lamins) and the conserved BAF protein that serves as a central link in these interactions, providing topological relationships between chromatin and nuclear envelope. The dynamic transformations of these protein ensembles in the mitotic cycle are characterized in detail at the molecular level, however, less attention is paid to the developing germ cells undergoing meiotic divisions, despite of their nuclei, especially in diplotene oocytes, differ significantly in structure from the somatic nucleus. This review summarizes the still relatively scarce experimental data proving the significance of functional interactions between BAF and LEMD proteins for gamete formation, from the selection of germline cells to the transformation of haploid spermatids into morphologically mature spermatozoa.

Insilico molecular modelling to identify PDK-1 targeting agents based on its protein–protein docking interaction

PDK1, an attractive cancer target that downstreams 23 other kinases towards cell growth, survival and metabolism has gaining attention due to allosteric effect of ligands bound to it. Generally, the drug design strategy using pharmacophores is either a single protein structure or ensemble or ligand-based. Apart from these methods, yet another new approach of protein–protein docking with state of art computational tool like Schrodinger Suite to generate pharmacophores based on the interacting partners of the protein is proposed in this work. The structure-based pharmacophoric features were picked up from docking the ten interacting partners of PDK1 and screened against the Enamine libraries containing protein–protein interacting compound collection, advanced, protein mimetic and allosteric compounds. High throughput virtual screening against the PIF pocket of PDK1 yields an indole scaffold. The identified indole derivative is proposed to be a strong activator that binds in the protein–protein interaction site of PDK1 which was further confirmed by molecular metadynamics simulations, free energy surface analysis and MM-GBSA calculations. Thus, the pharmacophores generated by the interacting proteins for PPI can facilitate the virtual screening in structure-based drug discovery of similar therapeutic targets. Communicated by Ramaswamy H. Sarma

The interplay of quaternary ammonium lipid structure and protein corona on lung-specific mRNA delivery by selective organ targeting (SORT) nanoparticles

Messenger RNA (mRNA) can treat genetic disease using protein replacement or genome editing approaches but requires a suitable carrier to circumnavigate biological barriers and access the desired cell type within the target organ. Lipid nanoparticles (LNPs) are widely used in the clinic for mRNA delivery yet are limited in their applications due to significant hepatic accumulation because of the formation of a protein corona enriched in apolipoprotein E (ApoE). Our lab developed selective organ targeting (SORT) LNPs that incorporate a supplementary component, termed a SORT molecule, for tissue-specific mRNA delivery to the liver, spleen, and lungs of mice. Mechanistic work revealed that the biophysical class of SORT molecule added to the LNP forms a distinct protein corona that helps determine where in the body mRNA is delivered. To better understand which plasma proteins could drive tissue-specific mRNA delivery, we characterized a panel of quaternary ammonium lipids as SORT molecules to assess how chemical structure affects the organ-targeting outcomes and protein corona of lung-targeting SORT LNPs. We discovered that variations in the chemical structure of both the lipid alkyl tail and headgroup impact the potency and specificity of mRNA delivery to the lungs. Furthermore, changes to the chemical structure alter the quantities and identities of protein corona constituents in a manner that correlates with organ-targeting outcomes, with certain proteins appearing to promote lung targeting whereas others reduce delivery to off-target organs. These findings unveil a nuanced relationship between LNP chemistry and endogenous targeting, where the ensemble of proteins associated with an LNP can play various roles in determining the tissue-specificity of mRNA delivery, providing further design criteria for optimization of clinically-relevant nanoparticles for extrahepatic delivery of genetic payloads.

Discovery of a Hidden Trypanosoma cruzi Spermidine Synthase Binding Site and Inhibitors through In Silico, In Vitro, and X-ray Crystallography.

In drug discovery research, the selection of promising binding sites and understanding the binding mode of compounds are crucial fundamental studies. The current understanding of the proteins-ligand binding model extends beyond the simple lock and key model to include the induced-fit model, which alters the conformation to match the shape of the ligand, and the pre-existing equilibrium model, selectively binding structures with high binding affinity from a diverse ensemble of proteins. Although methods for detecting target protein binding sites and virtual screening techniques using docking simulation are well-established, with numerous studies reported, they only consider a very limited number of structures in the diverse ensemble of proteins, as these methods are applied to a single structure. Molecular dynamics (MD) simulation is a method for predicting protein dynamics and can detect potential ensembles of protein binding sites and hidden sites unobservable in a single-point structure. In this study, to demonstrate the utility of virtual screening with protein dynamics, MD simulations were performed on Trypanosoma cruzi spermidine synthase to obtain an ensemble of dominant binding sites with a high probability of existence. The structure of the binding site obtained through MD simulation revealed pockets in addition to the active site that was present in the initial structure. Using the obtained binding site structures, virtual screening of 4.8 million compounds by docking simulation, in vitro assays, and X-ray analysis was conducted, successfully identifying two hit compounds.

PyInteraph2 and PyInKnife2 to Analyze Networks in Protein Structural Ensembles.

Due to the complex nature of noncovalent interactions and their long-range effects, analyzing protein conformations using network theory can be enlightening. Protein Structure Networks (PSNs) provide a convenient formalism to study protein structures in relation to essential properties such as key residues for structural stability, allosteric communication, and the effects of modifications of the protein. PSNs can be defined according to very different principles, and the available tools have limitations in input formats, supported models, and version control. Other outstanding problems are related to the definition of network cutoffs and the assessment of the stability of the network properties. The protein science community could benefit from a common framework to carry out these analyses and make them easier to reproduce, reuse, and evaluate. We here provide two open-source software packages, PyInteraph2 and PyInKnife2, to implement and analyze PSNs in a reproducible and documented manner. PyInteraph2 interfaces with multiple formats for protein ensembles and incorporates different network models with the possibility of integrating them into a macronetwork and performing various downstream analyses, including hubs, connected components, and several other centrality measures, and visualizes the networks or further analyzes them thanks to compatibility with Cytoscape.PyInKnife2 that supports the network models implemented in PyInteraph2. It employs a jackknife resampling approach to estimate the convergence of network properties and streamline the selection of distance cutoffs. We foresee that the modular structure of the code and the supported version control system will promote the transition to a community-driven effort, boost reproducibility, and establish common protocols in the PSN field. As developers, we will guarantee the introduction of new functionalities and maintenance, assistance, and training of new contributors.

Conformational Changes in Surface-Immobilized Proteins Measured Using Combined Atomic Force and Fluorescence Microscopy.

Biological organisms rely on proteins to perform the majority of their functions. Most protein functions are based on their physical motions (conformational changes), which can be described as transitions between different conformational states in a multidimensional free-energy landscape. A comprehensive understanding of this free-energy landscape is therefore of paramount importance for understanding the biological functions of proteins. Protein dynamics includes both equilibrium and nonequilibrium motions, which typically exhibit a wide range of characteristic length and time scales. The relative probabilities of various conformational states in the energy landscape, the energy barriers between them, their dependence on external parameters such as force and temperature, and their connection to the protein function remain largely unknown for most proteins. In this paper, we present a multimolecule approach in which the proteins are immobilized at well-defined locations on Au substrates using an atomic force microscope (AFM)-based patterning method called nanografting. This method enables precise control over the protein location and orientation on the substrate, as well as the creation of biologically active protein ensembles that self-assemble into well-defined nanoscale regions (protein patches) on the gold substrate. We performed AFM-force compression and fluorescence experiments on these protein patches and measured the fundamental dynamical parameters such as protein stiffness, elastic modulus, and transition energies between distinct conformational states. Our results provide new insights into the processes that govern protein dynamics and its connection to protein function.

PROTHON: A Local Order Parameter-Based Method for Efficient Comparison of Protein Ensembles.

The comparison of protein conformational ensembles is of central importance in structural biology. However, there are few computational methods for ensemble comparison, and those that are readily available, such as ENCORE, utilize methods that are sufficiently computationally expensive to be prohibitive for large ensembles. Here, a new method is presented for efficient representation and comparison of protein conformational ensembles. The method is based on the representation of a protein ensemble as a vector of probability distribution functions (pdfs), with each pdf representing the distribution of a local structural property such as the number of contacts between Cβ atoms. Dissimilarity between two conformational ensembles is quantified by the Jensen-Shannon distance between the corresponding set of probability distribution functions. The method is validated for conformational ensembles generated by molecular dynamics simulations of ubiquitin, as well as experimentally derived conformational ensembles of a 130 amino acid truncated form of human tau protein. In the ubiquitin ensemble data set, the method was up to 88 times faster than the existing ENCORE software, while simultaneously utilizing 48 times fewer computing cores. We make the method available as a Python package, called PROTHON, and provide a GitHub page with the Python source code at https://github.com/PlotkinLab/Prothon.

Machine Learning Generation of Dynamic Protein Conformational Ensembles.

Machine learning has achieved remarkable success across a broad range of scientific and engineering disciplines, particularly its use for predicting native protein structures from sequence information alone. However, biomolecules are inherently dynamic, and there is a pressing need for accurate predictions of dynamic structural ensembles across multiple functional levels. These problems range from the relatively well-defined task of predicting conformational dynamics around the native state of a protein, which traditional molecular dynamics (MD) simulations are particularly adept at handling, to generating large-scale conformational transitions connecting distinct functional states of structured proteins or numerous marginally stable states within the dynamic ensembles of intrinsically disordered proteins. Machine learning has been increasingly applied to learn low-dimensional representations of protein conformational spaces, which can then be used to drive additional MD sampling or directly generate novel conformations. These methods promise to greatly reduce the computational cost of generating dynamic protein ensembles, compared to traditional MD simulations. In this review, we examine recent progress in machine learning approaches towards generative modeling of dynamic protein ensembles and emphasize the crucial importance of integrating advances in machine learning, structural data, and physical principles to achieve these ambitious goals.

Structural Preferences Shape the Entropic Force of Disordered Protein Ensembles.

Intrinsically disordered protein regions (IDRs) make up over 30% of the human proteome and exist in a dynamic conformational ensemble instead of a native, well-folded structure. Tethering IDRs to a surface (for example, the surface of a well-folded region of the same protein) can reduce the number of accessible conformations in these ensembles. This reduces the ensemble's conformational entropy, generating an effective entropic force that pulls away from the point of tethering. Recent experimental work has shown that this entropic force causes measurable, physiologically relevant changes to protein function. But how the magnitude of this force depends on IDR sequence remains unexplored. Here, we use all-atom simulations to analyze how structural preferences in IDR ensembles contribute to the entropic force they exert upon tethering. We show that sequence-encoded structural preferences play an important role in determining the magnitude of this force: compact, spherical ensembles generate an entropic force that can be several times higher than more extended ensembles. We further show that changes in the surrounding solution's chemistry can modulate the IDR entropic force strength. We propose that the entropic force is a sequence-dependent, environmentally tunable property of terminal IDR sequences.

Active Learning of the Conformational Ensemble of Proteins Using Maximum Entropy VAMPNets.

Rapid computational exploration of the free energy landscape of biological molecules remains an active area of research due to the difficulty of sampling rare state transitions in molecular dynamics (MD) simulations. In recent years, an increasing number of studies have exploited machine learning (ML) models to enhance and analyze MD simulations. Notably, unsupervised models that extract kinetic information from a set of parallel trajectories have been proposed including the variational approach for Markov processes (VAMP), VAMPNets, and time-lagged variational autoencoders (TVAE). In this work, we propose a combination of adaptive sampling with active learning of kinetic models to accelerate the discovery of the conformational landscape of biomolecules. In particular, we introduce and compare several techniques that combine kinetic models with two adaptive sampling regimes (least counts and multiagent reinforcement learning-based adaptive sampling) to enhance the exploration of conformational ensembles without introducing biasing forces. Moreover, inspired by the active learning approach of uncertainty-based sampling, we also present MaxEnt VAMPNet. This technique consists of restarting simulations from the microstates that maximize the Shannon entropy of a VAMPNet trained to perform the soft discretization of metastable states. By running simulations on two test systems, the WLALL pentapeptide and the villin headpiece subdomain, we empirically demonstrate that MaxEnt VAMPNet results in faster exploration of conformational landscapes compared with the baseline and other proposed methods.

TMAO: Protecting proteins from feeling the heat

Osmolytes are ubiquitous in the cell and play an important role in controlling protein stability under stress. The natural osmolyte trimethylamine N-oxide (TMAO) is used by marine animals to counteract the effect of pressure denaturation at large depths. The molecular mechanism of TMAO stabilization against pressure and urea denaturation has been extensively studied, but unlike the case of other osmolytes, the ability of TMAO to protect proteins from high temperature has not been quantified. To reveal the effect of TMAO on folded and unfolded protein ensembles and the hydration shell at different temperatures, we study a mutant of the well-characterized, fast-folding model protein B (PRB). We carried out, in total, >190 μs all-atom simulations of thermal folding/unfolding of PRB at multiple temperatures and concentrations of TMAO. The simulations show increased thermal stability of PRB in the presence of TMAO. Partly structured, compact ensembles are favored over the unfolded state. TMAO forms two shells near the protein: an outer shell away from the protein surface has altered H-bond lifetimes of water molecules and increases hydration of the protein to help stabilize it; a less-populated inner shell with an opposite TMAO orientation closer to the protein surface binds exclusively to basic side chains. The cooperative cosolute effect of the inner and outer shell TMAO has a small number of TMAO molecules “herding” water molecules into two hydration shells at or near the protein surface. The stabilizing effect of TMAO on our protein saturates at 1 M despite higher TMAO solubility, so there may be little evolutionary pressure for extremophiles to produce higher intracellular TMAO concentrations, if true in general.

Multiplex protein analysis and ensemble machine learning methods of fine needle aspirates from prostate cancer patients reveal potential diagnostic signatures associated with tumour grade.

Improved molecular diagnosis is needed in prostate cancer (PC). Fine needle aspiration (FNA) is a minimally invasive biopsy technique, less traumatic compared to core needle biopsy, and could be useful for diagnosis of PC. Molecular biomarkers (BMs) in FNA-samples can be assessed for prediction, eg of immunotherapy efficacy before treatment as well as at treatment decision time points during disease progression. In the present pilot study, the expression levels of 151 BM proteins were analysed by proximity extension assay in FNA-samples from 16 patients, including benign prostate lesions (n= 3) and cancers (n= 13). An ensemble data analysis strategy was applied using several machine learning models. Twelve potentially predictive BM proteins correlating with International Society of Urological Pathology grade groups were identified, among them vimentin, tissue factor pathway inhibitor 2, and integrin beta-5. The validity of the results was supported by network analysis that showed functional associations between most of the identified putative BMs. We also showed that multiple immune checkpoint targets can be assessed (eg PD-L1, CD137, and Galectin-9), which may support the selection of immunotherapy in advanced PC. Results are promising but need further validation in a larger cohort. Our pilot study represents a "proof of concept" and shows that multiplex profiling of potential diagnostic and predictive BM proteins is feasible on tumour material obtained by FNA sampling of prostate cancer. Moreover, our results demonstrate that an ensemble data analysis strategy may facilitate the identification of BM signatures in pilot studies when the patient cohort is limited.

Generation of unfolded outer membrane protein ensembles defined by hydrodynamic properties.

Outer membrane proteins (OMPs) must exist as an unfolded ensemble while interacting with a chaperone network in the periplasm of Gram-negative bacteria. Here, we developed a method to model unfolded OMP (uOMP) conformational ensembles using the experimental properties of two well-studied OMPs. The overall sizes and shapes of the unfolded ensembles in the absence of a denaturant were experimentally defined by measuring the sedimentation coefficient as a function of urea concentration. We used these data to model a full range of unfolded conformations by parameterizing a targeted coarse-grained simulation protocol. The ensemble members were further refined by short molecular dynamics simulations to reflect proper torsion angles. The final conformational ensembles have polymer properties different from unfolded soluble and intrinsically disordered proteins and reveal inherent differences in the unfolded states that necessitate further investigation. Building these uOMP ensembles advances the understanding of OMP biogenesis and provides essential information for interpreting structures of uOMP-chaperone complexes.

Direct generation of protein conformational ensembles via machine learning

Dynamics and conformational sampling are essential for linking protein structure to biological function. While challenging to probe experimentally, computer simulations are widely used to describe protein dynamics, but at significant computational costs that continue to limit the systems that can be studied. Here, we demonstrate that machine learning can be trained with simulation data to directly generate physically realistic conformational ensembles of proteins without the need for any sampling and at negligible computational cost. As a proof-of-principle we train a generative adversarial network based on a transformer architecture with self-attention on coarse-grained simulations of intrinsically disordered peptides. The resulting model, idpGAN, can predict sequence-dependent coarse-grained ensembles for sequences that are not present in the training set demonstrating that transferability can be achieved beyond the limited training data. We also retrain idpGAN on atomistic simulation data to show that the approach can be extended in principle to higher-resolution conformational ensemble generation.