One bead per residue can describe all-atom protein structures

CABS-flex 3.0: an online tool for simulating protein structural flexibility and peptide modeling

Simulating protein structure flexibility using classical methods is computationally demanding, especially for large proteins. To address this challenge, we have been developing the CABS-flex method, which enables fast simulations of protein structural flexibility by combining a coarse-grained simulation approach with all-atom detail. Previously available as the CABS-flex 2.0 web server, the method has now undergone a major upgrade with the release of CABS-flex 3.0. Key improvements include the introduction of intuitive flexibility modes that simplify the control of distance restraints and allow users to reflect known or expected dynamic regions; improved all-atom reconstruction for higher-quality model generation; a new feature for de novo peptide structure prediction, supporting both linear and cyclic peptides along with their conformational flexibility; and new tools for result analysis and visualization, facilitating deeper insights into structural flexibility. Additionally, AlphaFold pLDDT-derived restraints can be used as optional input for guiding simulations. The method accepts input as either a PDB/mmCIF structure or a sequence (for peptide modeling). Advanced options allow users to incorporate experimental or computational restraints. The CABS-flex 3.0 web server is available at https://lcbio.pl/cabsflex3. This website is free and open to all users, with no login requirement.

Accurate Predictions of Molecular Properties of Proteins via Graph Neural Networks and Transfer Learning

Machine learning has emerged as a promising approach for predicting molecular properties of proteins, as it addresses limitations of experimental and traditional computational methods. Here, we introduce GSnet, a graph neural network (GNN) trained to predict physicochemical and geometric properties including solvation-free energies, diffusion constants, and hydrodynamic radii, based on three-dimensional protein structures. By leveraging transfer learning, pretrained GSnet embeddings were adapted to predict solvent-accessible surface area (SASA) and residue-specific pKa values, achieving high accuracy and generalizability. Notably, GSnet outperformed existing protein embeddings for SASA prediction and a locally charge-aware variant, aLCnet, approached the accuracy of simulation-based and empirical methods for pKa prediction. Our GNN framework demonstrated robustness across diverse data sets, including intrinsically disordered peptides, and scalability for high-throughput applications. These results highlight the potential of GNN-based embeddings and transfer learning to advance protein structure analysis, providing a foundation for integrating predictive models into proteome-wide studies and structural biology pipelines.

ART-SM: Boosting Fragment-Based Backmapping by Machine Learning

In sequential multiscale molecular dynamics simulations, which advantageously combine the increased sampling and dynamics at coarse-grained resolution with the higher accuracy of atomistic simulations, the resolution is altered over time. While coarse-graining is straightforward once the mapping between atomistic and coarse-grained resolution is defined, reintroducing the atomistic details is still a nontrivial process called backmapping. Here, we present ART-SM, a fragment-based backmapping framework that learns from atomistic simulation data to seamlessly switch from coarse-grained to atomistic resolution. ART-SM requires minimal user input and goes beyond state-of-the-art fragment-based approaches by selecting from multiple conformations per fragment via machine learning to simultaneously reflect the coarse-grained structure and the Boltzmann distribution. Additionally, we introduce a novel refinement step to connect individual fragments by optimizing specific bonds, angles, and dihedral angles in the backmapping process. We demonstrate that our algorithm accurately restores the atomistic bond length, angle, and dihedral angle distributions for various small and linear molecules from Martini coarse-grained beads and that the resulting high-resolution structures are representative of the input coarse-grained conformations. Moreover, the reconstruction of the TIP3P water model is fast and robust, and we demonstrate that ART-SM can be applied to larger linear molecules as well. To illustrate the efficiency of the local and autoregressive approach of ART-SM, we simulated a large realistic system containing the surfactants TAPB and SDS in solution using the Martini3 force field. The self-assembled micelles of various shapes were backmapped with ART-SM after training on only short atomistic simulations of a single water-solvated SDS or TAPB molecule. Together, these results indicate the potential for the method to be extended to more complex molecules such as lipids, proteins, macromolecules, and materials in the future.

Deep generative modeling of temperature-dependent structural ensembles of proteins

Deep learning has revolutionized protein structure prediction, but capturing conformational ensembles and structural variability remains an open challenge. While molecular dynamics (MD) is the foundation method for simulating biomolecular dynamics, it is computationally expensive. Recently, deep learning models trained on MD have made progress in generating structural ensembles at reduced cost. However, they remain limited in modeling atomistic details and, crucially, incorporating the effect of environmental factors. Here, we present aSAM (atomistic structural autoencoder model), a latent diffusion model trained on MD to generate heavy atom protein ensembles. Unlike most methods, aSAM models atoms in a latent space, greatly facilitating accurate sampling of side chain and backbone torsion angle distributions. Additionally, we extended aSAM into the first reported transferable generator conditioned on temperature, named aSAMt. Trained on the large and open mdCATH dataset, aSAMt captures temperature-dependent ensemble properties and demonstrates generalization beyond training temperatures. By comparing aSAMt ensembles to long MD simulations of fast folding proteins, we find that high-temperature training enhances the ability of deep generators to explore energy landscapes. Finally, we also show that our MD-based aSAMt can already capture experimentally observed thermal behavior of proteins. Our work is a step towards generalizable ensemble generation to complement physics-based approaches.

Machine learning methods to study sequence-ensemble-function relationships in disordered proteins

Recent years have seen tremendous developments in the use of machine learning models to link amino-acid sequence, structure, and function of folded proteins. These methods are, however, rarely applicable to the wide range of proteins and sequences that comprise intrinsically disordered regions. We here review developments in the study of sequence-ensemble-function relationships of disordered proteins that exploit or are used to train machine learning models. These include methods for generating conformational ensembles and designing new sequences, and for linking sequences to biophysical properties and biological functions. We highlight how these developments are built on a tight integration between experiment, theory and simulations, and account for evolutionary constraints, which operate on sequences of disordered regions differently than on those of folded domains.

COCOMO2: A Coarse-Grained Model for Interacting Folded and Disordered Proteins

Biomolecular interactions are essential in many biological processes, including complex formation and phase separation processes. Coarse-grained computational models are especially valuable for studying such processes via simulation. Here, we present COCOMO2, an updated residue-based coarse-grained model that extends its applicability from intrinsically disordered peptides to folded proteins. This is accomplished with the introduction of a surface exposure scaling factor, which adjusts interaction strengths based on solvent accessibility, to enable the more realistic modeling of interactions involving folded domains without additional computational costs. COCOMO2 was parametrized directly with solubility and phase separation data to improve its performance on predicting concentration-dependent phase separation for a broader range of biomolecular systems compared to the original version. COCOMO2 enables new applications including the study of condensates that involve IDPs together with folded domains and the study of complex assembly processes. COCOMO2 also provides an expanded foundation for the development of multiscale approaches for modeling biomolecular interactions that span from residue-level to atomistic resolution.

Implementation of Time-Averaged Restraints with UNRES Coarse-Grained Model of Polypeptide Chains

Time-averaged restraints from nuclear magnetic resonance (NMR) measurements have been implemented in the UNRES coarse-grained model of polypeptide chains in order to develop a tool for data-assisted modeling of the conformational ensembles of multistate proteins, intrinsically disordered proteins (IDPs) and proteins with intrinsically disordered regions (IDRs), many of which are essential in cell biology. A numerically stable variant of molecular dynamics with time-averaged restraints has been introduced, in which the total energy is conserved in sections of a trajectory in microcanonical runs, the bath temperature is maintained in canonical runs, and the time-average-restraint-force components are scaled up with the length of the memory window so that the restraints affect the simulated structures. The new approach restores the conformational ensembles used to generate ensemble-averaged distances, as demonstrated with synthetic restraints. The approach results in a better fitting of the ensemble-averaged interproton distances to those determined experimentally for multistate proteins and proteins with intrinsically disordered regions, which puts it at an advantage over all-atom approaches with regard to the determination of the conformational ensembles of proteins with diffuse structures, owing to a faster and more robust conformational search.

FlowBack: A Generalized Flow-Matching Approach for Biomolecular Backmapping

Coarse-grained models have become ubiquitous in biomolecular modeling tasks aimed at studying slow dynamical processes such as protein folding and DNA hybridization. These models can considerably accelerate sampling but it remains challenging to accurately and efficiently restore all-atom detail to the coarse-grained trajectory, which can be vital for detailed understanding of molecular mechanisms and calculation of observables contingent on all-atom coordinates. In this work, we introduce FlowBack as a deep generative model employing a flow-matching objective to map samples from a coarse-grained prior distribution to an all-atom data distribution. We construct our prior distribution to be agnostic to the coarse-grained map and molecular type. A protein-specific model trained on ∼65k structures from the Protein Data Bank achieves state-of-the-art performance on structural metrics compared to previous generative and rules-based approaches in applications to static PDB structures, all-atom simulations of fast-folding proteins, and coarse-grained trajectories generated by a machine-learned force field. A DNA-protein model trained on ∼1.5k DNA-protein complexes achieves excellent reconstruction and generative capabilities on static DNA-protein complexes from the Protein Data Bank as well as on out-of-distribution coarse-grained dynamical simulations of DNA-protein complexation. FlowBack offers an accurate, efficient, and easy-to-use tool to recover all-atom structures from coarse-grained molecular simulations with higher robustness and fewer steric clashes than previous approaches. We make FlowBack freely available to the community as an open source Python package.

Coarse-Grained Simulation Study of the Association of Selected Dipeptides

The association of 55 dipeptides extracted from aggregation-prone regions of selected proteins was studied by means of multiplexed replica-exchange molecular dynamics simulations with the coarse-grained UNRES model of polypeptide chains. Each simulation was carried out with 320 dipeptide molecules in a periodic box at 0.24 mol/dm3 concentration, in the 260-370 K temperature range. The temperature profiles of the degree of association, distributions of dipeptide cluster size, and structures of clusters were examined. It has been found that the dipeptides composed of strongly nonpolar (aromatic or aliphatic) residues associate nearly completely at all temperatures to form tight clusters, while those composed of charged or polar residues exhibited no or residual association. The dipeptides composed of nonpolar and small polar residues and those composed of less hydrophobic residues formed single clusters, gradually dissolving with increasing temperature, while those composed of phenylalanine or tryptophan and polar or charged residues formed multiple irregular clusters with room to accommodate water inside, suggesting the formation of liquid droplets or gels. The logarithms of the average degree of association and the free energy of aggregation per monomer were found to correlate with the dipeptide hydrophobicity.

Accurate Predictions of Molecular Properties of Proteins via Graph Neural Networks and Transfer Learning

Machine learning has emerged as a promising approach for predicting molecular properties of proteins, as it addresses limitations of experimental and traditional computational methods. Here, we introduce GSnet, a graph neural network (GNN) trained to predict physicochemical and geometric properties including solvation free energies, diffusion constants, and hydrodynamic radii, based on three-dimensional protein structures. By leveraging transfer learning, pre-trained GSnet embeddings were adapted to predict solvent-accessible surface area (SASA) and residue-specific pKa values, achieving high accuracy and generalizability. Notably, GSnet outperformed existing protein embeddings for SASA prediction, and a locally charge-aware variant, aLCnet, approached the accuracy of simulation-based and empirical methods for pKa prediction. Our GNN framework demonstrated robustness across diverse datasets, including intrinsically disordered peptides, and scalability for high-throughput applications. These results highlight the potential of GNN-based embeddings and transfer learning to advance protein structure analysis, providing a foundation for integrating predictive models into proteome-wide studies and structural biology pipelines.

deepBBQ: A Deep Learning Approach to the Protein Backbone Reconstruction

Coarse-grained models have provided researchers with greatly improved computational efficiency in modeling structures and dynamics of biomacromolecules, but, to be practically useful, they need fast and accurate conversion methods back to the all-atom representation. Reconstruction of atomic details may also be required in the case of some experimental methods, like electron microscopy, which may provide Cα-only structures. In this contribution, we present a new method for recovery of all backbone atom positions from just the Cα coordinates. Our approach, called deepBBQ, uses a deep convolutional neural network to predict a single internal coordinate per peptide plate, based on Cα trace geometric features, and then proceeds to recalculate the cartesian coordinates based on the assumption that the peptide plate atoms lie in the same plane. Extensive comparison with similar programs shows that our solution is accurate and cost-efficient. The deepBBQ program is available as part of the open-source bioinformatics toolkit Bioshell and is free for download and the documentation is available online.

Weighted families of contact maps to characterize conformational ensembles of (highly-)flexible proteins

MOTIVATION

Characterizing the structure of flexible proteins, particularly within the realm of intrinsic disorder, presents a formidable challenge due to their high conformational variability. Currently, their structural representation relies on (possibly large) conformational ensembles derived from a combination of experimental and computational methods. The detailed structural analysis of these ensembles is a difficult task, for which existing tools have limited effectiveness.

RESULTS

This study proposes an innovative extension of the concept of contact maps to the ensemble framework, incorporating the intrinsic probabilistic nature of disordered proteins. Within this framework, a conformational ensemble is characterized through a weighted family of contact maps. To achieve this, conformations are first described using a refined definition of contact that appropriately accounts for the geometry of the inter-residue interactions and the sequence context. Representative structural features of the ensemble naturally emerge from the subsequent clustering of the resulting contact-based descriptors. Importantly, transiently populated structural features are readily identified within large ensembles. The performance of the method is illustrated by several use cases and compared with other existing approaches, highlighting its superiority in capturing relevant structural features of highly flexible proteins.

AVAILABILITY AND IMPLEMENTATION

An open-source implementation of the method is provided together with an easy-to-use Jupyter notebook, available at https://gitlab.laas.fr/moma/WARIO.

COCOMO2: A coarse-grained model for interacting folded and disordered proteins

Biomolecular interactions are essential in many biological processes, including complex formation and phase separation processes. Coarse-grained computational models are especially valuable for studying such processes via simulation. Here, we present COCOMO2, an updated residue-based coarse-grained model that extends its applicability from intrinsically disordered peptides to folded proteins. This is accomplished with the introduction of a surface exposure scaling factor, which adjusts interaction strengths based on solvent accessibility, to enable the more realistic modeling of interactions involving folded domains without additional computational costs. COCOMO2 was parameterized directly with solubility and phase separation data to improve its performance on predicting concentration-dependent phase separation for a broader range of biomolecular systems compared to the original version. COCOMO2 enables new applications including the study of condensates that involve IDPs together with folded domains and the study of complex assembly processes. COCOMO2 also provides an expanded foundation for the development of multi-scale approaches for modeling biomolecular interactions that span from residue-level to atomistic resolution.

A coarse-grained model for disordered and multi-domain proteins

Many proteins contain more than one folded domain, and such modular multi-domain proteins help expand the functional repertoire of proteins. Because of their larger size and often substantial dynamics, it may be difficult to characterize the conformational ensembles of multi-domain proteins by simulations. Here, we present a coarse-grained model for multi-domain proteins that is both fast and provides an accurate description of the global conformational properties in solution. We show that the accuracy of a one-bead-per-residue coarse-grained model depends on how the interaction sites in the folded domains are represented. Specifically, we find excessive domain-domain interactions if the interaction sites are located at the position of the Cα atoms. We also show that if the interaction sites are located at the center of mass of the residue, we obtain good agreement between simulations and experiments across a wide range of proteins. We then optimize our previously described CALVADOS model using this center-of-mass representation, and validate the resulting model using independent data. Finally, we use our revised model to simulate phase separation of both disordered and multi-domain proteins, and to examine how the stability of folded domains may differ between the dilute and dense phases. Our results provide a starting point for understanding interactions between folded and disordered regions in proteins, and how these regions affect the propensity of proteins to self-associate and undergo phase separation.

CHARMM at 45: Enhancements in Accessibility, Functionality, and Speed

Since its inception nearly a half century ago, CHARMM has been playing a central role in computational biochemistry and biophysics. Commensurate with the developments in experimental research and advances in computer hardware, the range of methods and applicability of CHARMM have also grown. This review summarizes major developments that occurred after 2009 when the last review of CHARMM was published. They include the following: new faster simulation engines, accessible user interfaces for convenient workflows, and a vast array of simulation and analysis methods that encompass quantum mechanical, atomistic, and coarse-grained levels, as well as extensive coverage of force fields. In addition to providing the current snapshot of the CHARMM development, this review may serve as a starting point for exploring relevant theories and computational methods for tackling contemporary and emerging problems in biomolecular systems. CHARMM is freely available for academic and nonprofit research at https://academiccharmm.org/program.

Exploring protein functions from structural flexibility using CABS-flex modeling

Understanding protein function often necessitates characterizing the flexibility of protein structures. However, simulating protein flexibility poses significant challenges due to the complex dynamics of protein systems, requiring extensive computational resources and accurate modeling techniques. In response to these challenges, the CABS-flex method has been developed as an efficient modeling tool that combines coarse-grained simulations with all-atom detail. Available both as a web server and a standalone package, CABS-flex is dedicated to a wide range of users. The web server version offers an accessible interface for straightforward tasks, while the standalone command-line program is designed for advanced users, providing additional features, analytical tools, and support for handling large systems. This paper examines the application of CABS-flex across various structure-function studies, facilitating investigations into the interplay among protein structure, dynamics, and function in diverse research fields. We present an overview of the current status of the CABS-flex methodology, highlighting its recent advancements, practical applications, and forthcoming challenges.

Transferable deep generative modeling of intrinsically disordered protein conformations

Intrinsically disordered proteins have dynamic structures through which they play key biological roles. The elucidation of their conformational ensembles is a challenging problem requiring an integrated use of computational and experimental methods. Molecular simulations are a valuable computational strategy for constructing structural ensembles of disordered proteins but are highly resource-intensive. Recently, machine learning approaches based on deep generative models that learn from simulation data have emerged as an efficient alternative for generating structural ensembles. However, such methods currently suffer from limited transferability when modeling sequences and conformations absent in the training data. Here, we develop a novel generative model that achieves high levels of transferability for intrinsically disordered protein ensembles. The approach, named idpSAM, is a latent diffusion model based on transformer neural networks. It combines an autoencoder to learn a representation of protein geometry and a diffusion model to sample novel conformations in the encoded space. IdpSAM was trained on a large dataset of simulations of disordered protein regions performed with the ABSINTH implicit solvent model. Thanks to the expressiveness of its neural networks and its training stability, idpSAM faithfully captures 3D structural ensembles of test sequences with no similarity in the training set. Our study also demonstrates the potential for generating full conformational ensembles from datasets with limited sampling and underscores the importance of training set size for generalization. We believe that idpSAM represents a significant progress in transferable protein ensemble modeling through machine learning.

Physics-Informed Deep Learning Approach for Reintroducing Atomic Detail in Coarse-Grained Configurations of Multiple Poly(lactic acid) Stereoisomers

Multiscale modeling of complex molecular systems, such as macromolecules, encompasses methods that combine information from fine and coarse representations of molecules to capture material properties over a wide range of spatiotemporal scales. Being able to exchange information between different levels of resolution is essential for the effective transfer of this information. The inverse problem of reintroducing atomistic degrees of freedom in coarse-grained (CG) molecular configurations is particularly challenging as, from a mathematical point of view, it is an ill-posed problem; the forward mapping from the atomistic to the CG description is typically defined via a deterministic operator ("one-to-one" problem), whereas the reversed mapping from the CG to the atomistic model refers to creating one representative configuration out of many possible ones ("one-to-many" problem). Most of the backmapping methods proposed so far balance accuracy, efficiency, and general applicability. This is particularly important for macromolecular systems with different types of isomers, i.e., molecules that have the same molecular formula and sequence of bonded atoms (constitution) but differ in the three-dimensional configurations of their atoms in space. Here, we introduce a versatile deep learning approach for backmapping multicomponent CG macromolecules with chiral centers, trained to learn structural correlations between polymer configurations at the atomistic level and their corresponding CG descriptions. This method is intended to be simple and flexible while presenting a generic solution for resolution transformation. In addition, the method is aimed to respect the structural features of the molecule, such as local packing, capturing therefore the physical properties of the material. As an illustrative example, we apply the model on linear poly(lactic acid) (PLA) in melt, which is one of the most popular biodegradable polymers. The framework is tested on a number of model systems starting from homopolymer stereoisomers of PLA to copolymers with randomly placed chiral centers. The results demonstrate the efficiency and efficacy of the new approach.

Transferable deep generative modeling of intrinsically disordered protein conformations

Intrinsically disordered proteins have dynamic structures through which they play key biological roles. The elucidation of their conformational ensembles is a challenging problem requiring an integrated use of computational and experimental methods. Molecular simulations are a valuable computational strategy for constructing structural ensembles of disordered proteins but are highly resource-intensive. Recently, machine learning approaches based on deep generative models that learn from simulation data have emerged as an efficient alternative for generating structural ensembles. However, such methods currently suffer from limited transferability when modeling sequences and conformations absent in the training data. Here, we develop a novel generative model that achieves high levels of transferability for intrinsically disordered protein ensembles. The approach, named idpSAM, is a latent diffusion model based on transformer neural networks. It combines an autoencoder to learn a representation of protein geometry and a diffusion model to sample novel conformations in the encoded space. IdpSAM was trained on a large dataset of simulations of disordered protein regions performed with the ABSINTH implicit solvent model. Thanks to the expressiveness of its neural networks and its training stability, idpSAM faithfully captures 3D structural ensembles of test sequences with no similarity in the training set. Our study also demonstrates the potential for generating full conformational ensembles from datasets with limited sampling and underscores the importance of training set size for generalization. We believe that idpSAM represents a significant progress in transferable protein ensemble modeling through machine learning.

From simple to complex: Reconstructing all-atom structures from coarse-grained models using cg2all

In this issue of Structure, Heo and Feig present cg2all, a novel deep-learning model capable of efficiently predicting all-atom protein structures from coarse-grained (CG) representations. The model maintains high accuracy, even when the CG model is simplified to a single bead per residue, and has a number of promising applications.