Computational reconstruction of atomistic protein structures from coarse-grained models

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.

Structure prediction of linear and cyclic peptides using CABS-flex

The structural modeling of peptides can be a useful aid in the discovery of new drugs and a deeper understanding of the molecular mechanisms of life. Here we present a novel multiscale protocol for the structure prediction of linear and cyclic peptides. The protocol combines two main stages: coarse-grained simulations using the CABS-flex standalone package and an all-atom reconstruction-optimization process using the Modeller program. We evaluated the protocol on a set of linear peptides and two sets of cyclic peptides, with cyclization through the backbone and disulfide bonds. A comparison with other state-of-the-art tools (APPTEST, PEP-FOLD, ESMFold and AlphaFold implementation in ColabFold) shows that for most cases, AlphaFold offers the highest resolution. However, CABS-flex is competitive, particularly when it comes to short linear peptides. As demonstrated, the protocol performance can be further improved by combination with the residue-residue contact prediction method or more efficient scoring. The protocol is included in the CABS-flex standalone package along with online documentation to aid users in predicting the structure of peptides and mini-proteins.

Backmapping with Mapping and Isomeric Information

I present a powerful and flexible backmapping tool named Multiscale Simulation Tool (mstool) that converts a coarse-grained (CG) system into all-atom (AA) resolution and only requires AA to CG mapping and isomeric information (cis/trans/dihedral/chiral). The backmapping procedure includes two simple steps: (a) AA atoms are randomly placed near the corresponding CG beads according to the provided mapping scheme. (b) Energy minimization is performed with two modifications in the AA force field (FF). First, nonbonded interactions are replaced with cosine functions to ensure the numerical stability. Second, additional torsions are imposed to maintain the molecules' isomeric properties. To test the simplicity and robustness of the tool, I backmapped multiple membrane and protein CG structures into AA resolution, including a four-bead CG lipid model (resolution increased by a factor of 34) without using intermediate resolution. The tool is freely available at github.com/ksy141/mstool.

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.

An end-to-end deep learning method for protein side-chain packing and inverse folding

Protein side-chain packing (PSCP), the task of determining amino acid side-chain conformations given only backbone atom positions, has important applications to protein structure prediction, refinement, and design. Many methods have been proposed to tackle this problem, but their speed or accuracy is still unsatisfactory. To address this, we present AttnPacker, a deep learning (DL) method for directly predicting protein side-chain coordinates. Unlike existing methods, AttnPacker directly incorporates backbone 3D geometry to simultaneously compute all side-chain coordinates without delegating to a discrete rotamer library or performing expensive conformational search and sampling steps. This enables a significant increase in computational efficiency, decreasing inference time by over 100× compared to the DL-based method DLPacker and physics-based RosettaPacker. Tested on the CASP13 and CASP14 native and nonnative protein backbones, AttnPacker computes physically realistic side-chain conformations, reducing steric clashes and improving both rmsd and dihedral accuracy compared to state-of-the-art methods SCWRL4, FASPR, RosettaPacker, and DLPacker. Different from traditional PSCP approaches, AttnPacker can also codesign sequences and side chains, producing designs with subnative Rosetta energy and high in silico consistency.

Quality assessment of VHH models

Heavy Chain Only Antibodies are specific to Camelid species. Despite the lack of the light chain variable domain, their heavy chain variable domain (VH) domain, named VHH or nanobody, has promising potential applications in research and therapeutic fields. The structural study of VHH is therefore of great interest. Unfortunately, considering the huge amount of sequences that might be produced, only about one thousand of VHH experimental structures are publicly available in the Protein Data Bank, implying that structural model prediction of VHH is a necessary alternative to obtaining 3D information besides its sequence. The present study aims to assess and compare the quality of predictions from different modelling methodologies. Established comparative & homology modelling approaches to recent Deep Learning-based modelling strategies were applied, i.e. Modeller using single or multiple structural templates, ModWeb, SwissModel (with two evaluation schema), RoseTTAfold, AlphaFold 2 and NanoNet. The prediction accuracy was evaluated using RMSD, TM-score, GDT-TS, GDT-HA and Protein Blocks distance metrics. Besides the global structure assessment, we performed specific analyses of Frameworks and CDRs structures. We observed that AlphaFold 2 and especially NanoNet performed better than the other evaluated softwares. Importantly, we performed molecular dynamics simulations of an experimental structure and a NanoNet predicted model of a VHH in order to compare the global structural flexibility and local conformations using Protein Blocks. Despite rather similar structures, substantial differences in dynamical properties were observed, which underlies the complexity of the task of model evaluation.Communicated by Ramaswamy H. Sarma.

Computational analysis of protein stability and allosteric interaction networks in distinct conformational forms of the SARS-CoV-2 spike D614G mutant: reconciling functional mechanisms through allosteric model of spike regulation

In this study, we used an integrative computational approach to examine molecular mechanisms underlying functional effects of the D614G mutation by exploring atomistic modeling of the SARS-CoV-2 spike proteins as allosteric regulatory machines. We combined coarse-grained simulations, protein stability and dynamic fluctuation communication analysis with network-based community analysis to examine structures of the native and mutant SARS-CoV-2 spike proteins in different functional states. Through distance fluctuations communication analysis, we probed stability and allosteric communication propensities of protein residues in the native and mutant SARS-CoV-2 spike proteins, providing evidence that the D614G mutation can enhance long-range signaling of the allosteric spike engine. By combining functional dynamics analysis and ensemble-based alanine scanning of the SARS-CoV-2 spike proteins we found that the D614G mutation can improve stability of the spike protein in both closed and open forms, but shifting thermodynamic preferences towards the open mutant form. Our results revealed that the D614G mutation can promote the increased number of stable communities and allosteric hub centers in the open form by reorganizing and enhancing the stability of the S1-S2 inter-domain interactions and restricting mobility of the S1 regions. This study provides atomistic-based view of allosteric communications in the SARS-CoV-2 spike proteins, suggesting that the D614G mutation can exert its primary effect through allosterically induced changes on stability and communications in the residue interaction networks.Communicated by Ramaswamy H. Sarma.

Deep convolutional neural networks for generating atomistic configurations of multi-component macromolecules from coarse-grained models

Despite the modern advances in the available computational resources, the length and time scales of the physical systems that can be studied in full atomic detail, via molecular simulations, are still limited. To overcome such limitations, coarse-grained (CG) models have been developed to reduce the dimensionality of the physical system under study. However, to study such systems at the atomic level, it is necessary to re-introduce the atomistic details into the CG description. Such an ill-posed mathematical problem is typically treated via numerical algorithms, which need to balance accuracy, efficiency, and general applicability. Here, we introduce an efficient and versatile method for backmapping multi-component CG macromolecules of arbitrary microstructures. By utilizing deep learning algorithms, we train a convolutional neural network to learn structural correlations between polymer configurations at the atomistic and their corresponding CG descriptions, obtained from atomistic simulations. The trained model is then utilized to get predictions of atomistic structures from input CG configurations. As an illustrative example, we apply the convolutional neural network to polybutadiene copolymers of various microstructures, in which each monomer microstructure (i.e., cis-1,4, trans-1,4, and vinyl-1,2) is represented as a different CG particle type. The proposed methodology is transferable over molecular weight and various microstructures. Moreover, starting from a specific single CG configuration with a given microstructure, we show that by modifying its chemistry (i.e., CG particle types), we are able to obtain a set of well equilibrated polymer configurations of different microstructures (chemistry) than the one of the original CG configuration.

Bottom-up Coarse-Graining: Principles and Perspectives

Large-scale computational molecular models provide scientists a means to investigate the effect of microscopic details on emergent mesoscopic behavior. Elucidating the relationship between variations on the molecular scale and macroscopic observable properties facilitates an understanding of the molecular interactions driving the properties of real world materials and complex systems (e.g., those found in biology, chemistry, and materials science). As a result, discovering an explicit, systematic connection between microscopic nature and emergent mesoscopic behavior is a fundamental goal for this type of investigation. The molecular forces critical to driving the behavior of complex heterogeneous systems are often unclear. More problematically, simulations of representative model systems are often prohibitively expensive from both spatial and temporal perspectives, impeding straightforward investigations over possible hypotheses characterizing molecular behavior. While the reduction in resolution of a study, such as moving from an atomistic simulation to that of the resolution of large coarse-grained (CG) groups of atoms, can partially ameliorate the cost of individual simulations, the relationship between the proposed microscopic details and this intermediate resolution is nontrivial and presents new obstacles to study. Small portions of these complex systems can be realistically simulated. Alone, these smaller simulations likely do not provide insight into collectively emergent behavior. However, by proposing that the driving forces in both smaller and larger systems (containing many related copies of the smaller system) have an explicit connection, systematic bottom-up CG techniques can be used to transfer CG hypotheses discovered using a smaller scale system to a larger system of primary interest. The proposed connection between different CG systems is prescribed by (i) the CG representation (mapping) and (ii) the functional form and parameters used to represent the CG energetics, which approximate potentials of mean force (PMFs). As a result, the design of CG methods that facilitate a variety of physically relevant representations, approximations, and force fields is critical to moving the frontier of systematic CG forward. Crucially, the proposed connection between the system used for parametrization and the system of interest is orthogonal to the optimization used to approximate the potential of mean force present in all systematic CG methods. The empirical efficacy of machine learning techniques on a variety of tasks provides strong motivation to consider these approaches for approximating the PMF and analyzing these approximations.

Asynchronous Reciprocal Coupling of Martini 2.2 Coarse-Grained and CHARMM36 All-Atom Simulations in an Automated Multiscale Framework

The appeal of multiscale modeling approaches is predicated on the promise of combinatorial synergy. However, this promise can only be realized when distinct scales are combined with reciprocal consistency. Here, we consider multiscale molecular dynamics (MD) simulations that combine the accuracy and macromolecular flexibility accessible to fixed-charge all-atom (AA) representations with the sampling speed accessible to reductive, coarse-grained (CG) representations. AA-to-CG conversions are relatively straightforward because deterministic routines with unique outcomes are achievable. Conversely, CG-to-AA conversions have many solutions due to a surge in the number of degrees of freedom. While automated tools for biomolecular CG-to-AA transformation exist, we find that one popular option, called Backward, is prone to stochastic failure and the AA models that it does generate frequently have compromised protein structure and incorrect stereochemistry. Although these shortcomings can likely be circumvented by human intervention in isolated instances, automated multiscale coupling requires reliable and robust scale conversion. Here, we detail an extension to Multiscale Machine-learned Modeling Infrastructure (MuMMI), including an improved CG-to-AA conversion tool called sinceCG. This tool is reliable (∼98% weakly correlated repeat success rate), automatable (no unrecoverable hangs), and yields AA models that generally preserve protein secondary structure and maintain correct stereochemistry. We describe how the MuMMI framework identifies CG system configurations of interest, converts them to AA representations, and simulates them at the AA scale while on-the-fly analyses provide feedback to update CG parameters. Application to systems containing the peripheral membrane protein RAS and proximal components of RAF kinase on complex eight-component lipid bilayers with ∼1.5 million atoms is discussed in the context of MuMMI.

DLPacker: Deep learning for prediction of amino acid side chain conformations in proteins

Prediction of side chain conformations of amino acids in proteins (also termed "packing") is an important and challenging part of protein structure prediction with many interesting applications in protein design. A variety of methods for packing have been developed but more accurate ones are still needed. Machine learning (ML) methods have recently become a powerful tool for solving various problems in diverse areas of science, including structural biology. In this study, we evaluate the potential of deep neural networks (DNNs) for prediction of amino acid side chain conformations. We formulate the problem as image-to-image transformation and train a U-net style DNN to solve the problem. We show that our method outperforms other physics-based methods by a significant margin: reconstruction RMSDs for most amino acids are about 20% smaller compared to SCWRL4 and Rosetta Packer with RMSDs for bulky hydrophobic amino acids Phe, Tyr, and Trp being up to 50% smaller.

Implementation of residue-level coarse-grained models in GENESIS for large-scale molecular dynamics simulations

Residue-level coarse-grained (CG) models have become one of the most popular tools in biomolecular simulations in the trade-off between modeling accuracy and computational efficiency. To investigate large-scale biological phenomena in molecular dynamics (MD) simulations with CG models, unified treatments of proteins and nucleic acids, as well as efficient parallel computations, are indispensable. In the GENESIS MD software, we implement several residue-level CG models, covering structure-based and context-based potentials for both well-folded biomolecules and intrinsically disordered regions. An amino acid residue in protein is represented as a single CG particle centered at the Cα atom position, while a nucleotide in RNA or DNA is modeled with three beads. Then, a single CG particle represents around ten heavy atoms in both proteins and nucleic acids. The input data in CG MD simulations are treated as GROMACS-style input files generated from a newly developed toolbox, GENESIS-CG-tool. To optimize the performance in CG MD simulations, we utilize multiple neighbor lists, each of which is attached to a different nonbonded interaction potential in the cell-linked list method. We found that random number generations for Gaussian distributions in the Langevin thermostat are one of the bottlenecks in CG MD simulations. Therefore, we parallelize the computations with message-passing-interface (MPI) to improve the performance on PC clusters or supercomputers. We simulate Herpes simplex virus (HSV) type 2 B-capsid and chromatin models containing more than 1,000 nucleosomes in GENESIS as examples of large-scale biomolecular simulations with residue-level CG models. This framework extends accessible spatial and temporal scales by multi-scale simulations to study biologically relevant phenomena, such as genome-scale chromatin folding or phase-separated membrane-less condensations.

Computational models for studying physical instabilities in high concentration biotherapeutic formulations

Computational prediction of the behavior of concentrated protein solutions is particularly advantageous in early development stages of biotherapeutics when material availability is limited and a large set of formulation conditions needs to be explored. This review provides an overview of the different computational paradigms that have been successfully used in modeling undesirable physical behaviors of protein solutions with a particular emphasis on high-concentration drug formulations. This includes models ranging from all-atom simulations, coarse-grained representations to macro-scale mathematical descriptions used to study physical instability phenomena of protein solutions such as aggregation, elevated viscosity, and phase separation. These models are compared and summarized in the context of the physical processes and their underlying assumptions and limitations. A detailed analysis is also given for identifying protein interaction processes that are explicitly or implicitly considered in the different modeling approaches and particularly their relations to various formulation parameters. Lastly, many of the shortcomings of existing computational models are discussed, providing perspectives and possible directions toward an efficient computational framework for designing effective protein formulations.

Molecular simulation of linear octacosane via a CG10 coarse grain scheme

Following our previous work on the united-atom simulation on octacosane (C₂₈H₅₈) (Dai et al., Phys. Chem. Chem. Phys., 2021, 23, 21262-21271), we developed a coarse grain scheme (CG10), which is able to reproduce the pivotal phase characteristics of octacosane with highly improved computational efficiency. The CG10 octacosane chain was composed of 10 consecutive beads, maintaining the fundamental zigzag chain morphology. When the potential functions were set up and the coefficients were parameterized, our CG10 models yielded solid phase diagrams and transitions during an annealing process. We also detected the melting point by various means: direct observation, bond order, density tracking, and an enthalpy plot. Furthermore, our CG10 successfully reproduced the liquid density with only 2% underestimation, indicating its applicability across the solid and liquid phases. Therefore, with the ability to reproduce critical structure and property characteristics, our CG10 scheme provides an effective means of numerically modelling octacosane with highly improved computational efficiency.

Prediction of structural alphabet protein blocks using data mining

3D protein structures determine proteins' biological functions. The 3D structure of the protein backbone can be approximated using the prototypes of local protein conformations. Sets of these prototypes are called structural alphabets (SAs). Amongst several approaches to the prediction of 3D structures from amino acid sequences, one approach is based on the prediction of SA prototypes for a given amino acid sequence. Protein Blocks (PBs) is the most known SA, and it is composed of 16 prototypes of five consecutive amino acids which were identified as optimal prototypes considering the ability to correctly approximate the local structure and the prediction accuracy of prototypes from an amino acid sequence. We developed models for PBs prediction from sequence information using different data mining approaches and machine learning algorithms. Besides the amino acid sequences, the results of the following tools were used to train the models: the Spider3 predictor of protein structure properties, several predictors of the protein's intrinsically disordered regions, and a tool for finding repeats in amino acid sequences. The highest accuracy of the constructed models is 80%, which is a significant improvement compared to the previous best available prediction, whose accuracy was 61%. Analyzing the models constructed by applying different algorithms, it was noticed that the significance of input attributes differs among the models constructed by algorithms. Using the information about amino acids belonging to intrinsically disordered regions and repeats improves the precision of prediction for some PBs using the CART classification algorithm, while this is not the case with the C5.0 classification algorithm. Improved prediction approaches can have interesting applications in protein structural model approaches or computational protein design.

OUP accepted manuscript

Correct identification and effective visualization of interactions in biomolecular structures facilitate understanding of their functions and molecular design. In response to the practical needs of structure-based analysis, we have created a Mapiya web server. The Mapiya integrates four main functionalities: (i) generation of contact maps - intramolecular and intermolecular-for proteins, nucleic acids, and their complexes; (ii) characterization of the interactions physicochemical nature, (iii) interactive visualization of biomolecular conformations with automatic zoom on selected contacts using Molstar and (iv) additional sequence- and structure-based analyses performed with third-party software and in-house algorithms combined into an easy-to-use interface. Thus, Mapiya offers a highly customized analysis of the molecular interactions' in various biological systems. The web server is available at: http://mapiya.lcbio.pl/.

GLIMPS: A Machine Learning Approach to Resolution Transformation for Multiscale Modeling

We describe a general approach to transforming molecular models between different levels of resolution, based on machine learning methods. The approach uses a matched set of models at both levels of resolution for training, but requires only the coordinates of their particles and no side information (e.g., templates for substructures, defined mappings, or molecular mechanics force fields). Once trained, the approach can transform further molecular models of the system between the two levels of resolution in either direction with equal facility.

A conserved Neurite Outgrowth and Guidance motif with biomimetic potential in neuronal Cell Adhesion Molecules

The discovery of conserved protein motifs can, in turn, unveil important regulatory signals, and when properly designed, synthetic peptides derived from such motifs can be used as biomimetics for biotechnological and therapeutic purposes. We report here that specific Ig-like repeats from the extracellular domains of neuronal Cell Adhesion Molecules share a highly conserved Neurite Outgrowth and Guidance (NOG) motif, which mediates homo- and heterophilic interactions crucial in neural development and repair. Synthetic peptides derived from the NOG motif of such proteins can boost neuritogenesis, and this potential is also retained by peptides with recombinant sequences, when fitting the NOG sequence pattern. The NOG motif discovery not only provides one more tile to the complex puzzle of neuritogenesis, but also opens the route to new neural regeneration strategies via a tunable biomimetic toolbox.

Kinematic Reconstruction of Cyclic Peptides and Protein Backbones from Partial Data

We present an algorithm, QBKR (Quaternary Backbone Kinematic Reconstruction), a fast analytical method for an all-atom backbone reconstruction of proteins and linear or cyclic peptide chains from C_α coordinate traces. Unlike previous analytical methods for deriving all-atom representations from coarse-grained models that rely on canonical geometry with planar peptides in the trans conformation, our de novo kinematic model incorporates noncanonical, cis-trans, geometry naturally. Perturbations to this geometry can be effected with ease in our formulation, for example, to account for a continuous change from cis to trans geometry. A simple optimization of a spring-based objective function is employed for C_α-C_α distance variations that extend beyond the cis-trans limit. The kinematic construction produces a linked chain of peptide units, C_α-C-N-C_α, hinged at the C_α atoms spanning all possible planar and nonplanar peptide conformations. We have combined our method with a ring closure algorithm for the case of ring peptides and missing loops in a protein structure. Here, the reconstruction proceeding from both the N and C termini of the protein backbone (or in both directions from a starting position for rings) requires freedom in the position of one C_α atom (a capstone) to achieve a successful loop or ring closure. A salient feature of our reconstruction method is the ability to enrich conformational ensembles to produce alternative feasible conformations in which H-bond forming C-O or N-H pairs in the backbone can reverse orientations, thus addressing a well-known shortcoming in C_α-based RMSD structure comparison, wherein very close structures may lead to significantly different overall H-bond behavior. We apply the fixed C_α-based design to the reverse reconstruction from noisy Cryo-EM data, a posteriori to the optimization. Our method can be applied to speed up the process of an all-atom description from voluminous experimental data or subpar electron density maps.

Computational strategies for protein conformational ensemble detection

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

Rosetta:MSF:NN: Boosting performance of multi-state computational protein design with a neural network

Rational protein design aims at the targeted modification of existing proteins. To reach this goal, software suites like Rosetta propose sequences to introduce the desired properties. Challenging design problems necessitate the representation of a protein by means of a structural ensemble. Thus, Rosetta multi-state design (MSD) protocols have been developed wherein each state represents one protein conformation. Computational demands of MSD protocols are high, because for each of the candidate sequences a costly three-dimensional (3D) model has to be created and assessed for all states. Each of these scores contributes one data point to a complex, design-specific energy landscape. As neural networks (NN) proved well-suited to learn such solution spaces, we integrated one into the framework Rosetta:MSF instead of the so far used genetic algorithm with the aim to reduce computational costs. As its predecessor, Rosetta:MSF:NN administers a set of candidate sequences and their scores and scans sequence space iteratively. During each iteration, the union of all candidate sequences and their Rosetta scores are used to re-train NNs that possess a design-specific architecture. The enormous speed of the NNs allows an extensive assessment of alternative sequences, which are ranked on the scores predicted by the NN. Costly 3D models are computed only for a small fraction of best-scoring sequences; these and the corresponding 3D-based scores replace half of the candidate sequences during each iteration. The analysis of two sets of candidate sequences generated for a specific design problem by means of a genetic algorithm confirmed that the NN predicted 3D-based scores quite well; the Pearson correlation coefficient was at least 0.95. Applying Rosetta:MSF:NN:enzdes to a benchmark consisting of 16 ligand-binding problems showed that this protocol converges ten-times faster than the genetic algorithm and finds sequences with comparable scores.

Computational Tools to Rationalize and Predict the Self-Assembly Behavior of Supramolecular Gels

Supramolecular gels form a class of soft materials that has been heavily explored by the chemical community in the past 20 years. While a multitude of experimental techniques has demonstrated its usefulness when characterizing these materials, the potential value of computational techniques has received much less attention. This review aims to provide a complete overview of studies that employ computational tools to obtain a better fundamental understanding of the self-assembly behavior of supramolecular gels or to accelerate their development by means of prediction. As such, we hope to stimulate researchers to consider using computational tools when investigating these intriguing materials. In the concluding remarks, we address future challenges faced by the field and formulate our vision on how computational methods could help overcoming them.

Protein-Protein Docking with Large-Scale Backbone Flexibility Using Coarse-Grained Monte-Carlo Simulations

Most of the protein–protein docking methods treat proteins as almost rigid objects. Only the side-chains flexibility is usually taken into account. The few approaches enabling docking with a flexible backbone typically work in two steps, in which the search for protein–protein orientations and structure flexibility are simulated separately. In this work, we propose a new straightforward approach for docking sampling. It consists of a single simulation step during which a protein undergoes large-scale backbone rearrangements, rotations, and translations. Simultaneously, the other protein exhibits small backbone fluctuations. Such extensive sampling was possible using the CABS coarse-grained protein model and Replica Exchange Monte Carlo dynamics at a reasonable computational cost. In our proof-of-concept simulations of 62 protein–protein complexes, we obtained acceptable quality models for a significant number of cases.

Efficient Flexible Fitting Refinement with Automatic Error Fixing for De Novo Structure Modeling from Cryo-EM Density Maps

Structural modeling of proteins from cryo-electron microscopy (cryo-EM) density maps is one of the challenging issues in structural biology. De novo modeling combined with flexible fitting refinement (FFR) has been widely used to build a structure of new proteins. In de novo prediction, artificial conformations containing local structural errors such as chirality errors, cis peptide bonds, and ring penetrations are frequently generated and cannot be easily removed in the subsequent FFR. Moreover, refinement can be significantly suppressed due to the low mobility of atoms inside the protein. To overcome these problems, we propose an efficient scheme for FFR, in which the local structural errors are fixed first, followed by FFR using an iterative simulated annealing (SA) molecular dynamics protocol with the united atom (UA) model in an implicit solvent model; we call this scheme "SAUA-FFR". The best model is selected from multiple flexible fitting runs with various biasing force constants to reduce overfitting. We apply our scheme to the decoys obtained from MAINMAST and demonstrate an improvement of the best model of eight selected proteins in terms of the root-mean-square deviation, MolProbity score, and RWplus score compared to the original scheme of MAINMAST. Fixing the local structural errors can enhance the formation of secondary structures, and the UA model enables progressive refinement compared to the all-atom model owing to its high mobility in the implicit solvent. The SAUA-FFR scheme realizes efficient and accurate protein structure modeling from medium-resolution maps with less overfitting.

Molecular Dynamics Scoring of Protein-Peptide Models Derived from Coarse-Grained Docking

One of the major challenges in the computational prediction of protein-peptide complexes is the scoring of predicted models. Usually, it is very difficult to find the most accurate solutions out of the vast number of sometimes very different and potentially plausible predictions. In this work, we tested the protocol for Molecular Dynamics (MD)-based scoring of protein-peptide complex models obtained from coarse-grained (CG) docking simulations. In the first step of the scoring procedure, all models generated by CABS-dock were reconstructed starting from their original C-alpha trace representations to all-atom (AA) structures. The second step included geometry optimization of the reconstructed complexes followed by model scoring based on receptor-ligand interaction energy estimated from short MD simulations in explicit water. We used two well-known AA MD force fields, CHARMM and AMBER, and a CG MARTINI force field. Scoring results for 66 different protein-peptide complexes show that the proposed MD-based scoring approach can be used to identify protein-peptide models of high accuracy. The results also indicate that the scoring accuracy may be significantly affected by the quality of the reconstructed protein receptor structures.

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

To greatly expand the druggable genome, fast and accurate predictions of cryptic sites for small molecules binding in target proteins are in high demand. In this study, we have developed a fast and simple conformational sampling scheme guided by normal modes solved from the coarse-grained elastic models followed by atomistic backbone refinement and side-chain repacking. Despite the observations of complex and diverse conformational changes associated with ligand binding, we found that simply sampling along each of the lowest 30 modes is near optimal for adequately restructuring cryptic sites so they can be detected by existing pocket finding programs like fpocket and concavity. We further trained machine-learning protocols to optimize the combination of the sampling-enhanced pocket scores with other dynamic and conservation scores, which only slightly improved the performance. As assessed based on a training set of 84 known cryptic sites and a test set of 14 proteins, our method achieved high accuracy of prediction (with area under the receiver operating characteristic curve >0.8) comparable to the CryptoSite server. Compared with CryptoSite and other methods based on extensive molecular dynamics simulation, our method is much faster (1-2 hours for an average-size protein) and simpler (using only pocket scores), so it is suitable for high-throughput processing of large datasets of protein structures at the genome scale.

Protocols for Fast Simulations of Protein Structure Flexibility Using CABS-Flex and SURPASS

Conformational flexibility of protein structures can play an important role in protein function. The flexibility is often studied using computational methods since experimental characterization can be difficult. Depending on protein system size, computational tools may require large computational resources or significant simplifications in the modeled systems to speed up calculations. In this work, we present the protocols for efficient simulations of flexibility of folded protein structures that use coarse-grained simulation tools of different resolutions: medium, represented by CABS-flex, and low, represented by SUPRASS. We test the protocols using a set of 140 globular proteins and compare the results with structure fluctuations observed in MD simulations, ENM modeling, and NMR ensembles. As demonstrated, CABS-flex predictions show high correlation to experimental and MD simulation data, while SURPASS is less accurate but promising in terms of future developments.

DeepTracer for fast de novo cryo-EM protein structure modeling and special studies on CoV-related complexes

Electron cryomicroscopy (cryo-EM), a 2017 Nobel prize-awarded technology, provides direct 3D maps of macromolecules and explains the shape and interactions of protein complexes such as SARS-CoV-2 viral proteins and human cell receptors. This understanding can be combined with detailed structural information gathered using other technologies to form the basis for modeling course of diseases and for designing therapeutic drugs. However, ab initio modeling of protein complex structure remains a challenging problem. Here, we present DeepTracer, a fully automated and robust tool that determines the all-atom structure of a protein complex based solely on its cryo-EM map and amino acid sequence, with improved accuracy and efficiency compared to previous methods. We also provide a web service for global access.

Information about macromolecular structure of protein complexes and related cellular and molecular mechanisms can assist the search for vaccines and drug development processes. To obtain such structural information, we present DeepTracer, a fully automated deep learning-based method for fast de novo multichain protein complex structure determination from high-resolution cryoelectron microscopy (cryo-EM) maps. We applied DeepTracer on a previously published set of 476 raw experimental cryo-EM maps and compared the results with a current state of the art method. The residue coverage increased by over 30% using DeepTracer, and the rmsd value improved from 1.29 Å to 1.18 Å. Additionally, we applied DeepTracer on a set of 62 coronavirus-related cryo-EM maps, among them 10 with no deposited structure available in EMDataResource. We observed an average residue match of 84% with the deposited structures and an average rmsd of 0.93 Å. Additional tests with related methods further exemplify DeepTracer’s competitive accuracy and efficiency of structure modeling. DeepTracer allows for exceptionally fast computations, making it possible to trace around 60,000 residues in 350 chains within only 2 h. The web service is globally accessible at https://deeptracer.uw.edu.

Docking of peptides to GPCRs using a combination of CABS-dock with FlexPepDock refinement

The structural description of peptide ligands bound to G protein-coupled receptors (GPCRs) is important for the discovery of new drugs and deeper understanding of the molecular mechanisms of life. Here we describe a three-stage protocol for the molecular docking of peptides to GPCRs using a set of different programs: (1) CABS-dock for docking fully flexible peptides; (2) PD2 method for the reconstruction of atomistic structures from C-alpha traces provided by CABS-dock and (3) Rosetta FlexPepDock for the refinement of protein–peptide complex structures and model scoring. We evaluated the proposed protocol on the set of seven different GPCR–peptide complexes (including one containing a cyclic peptide), for which crystallographic structures are available. We show that CABS-dock produces high resolution models in the sets of top-scored models. These sets of models, after reconstruction to all-atom representation, can be further improved by Rosetta high-resolution refinement and/or minimization, leading in most of the cases to sub-Angstrom accuracy in terms of interface root-mean-square-deviation measure.

Coarse-grained (hybrid) integrative modeling of biomolecular interactions

The computational modeling field has vastly evolved over the past decades. The early developments of simplified protein systems represented a stepping stone towards establishing more efficient approaches to sample intricated conformational landscapes. Downscaling the level of resolution of biomolecules to coarser representations allows for studying protein structure, dynamics and interactions that are not accessible by classical atomistic approaches. The combination of different resolutions, namely hybrid modeling, has also been proved as an alternative when mixed levels of details are required. In this review, we provide an overview of coarse-grained/hybrid models focusing on their applicability in the modeling of biomolecular interactions. We give a detailed list of ready-to-use modeling software for studying biomolecular interactions allowing various levels of coarse-graining and provide examples of complexes determined by integrative coarse-grained/hybrid approaches in combination with experimental information.

Docking interactions determine early cleavage events in insulin proteolysis by pepsin: Experiment and simulation

In silico modelling of cascade enzymatic proteolysis is an exceedingly complex and challenging task. Here, we study partial proteolysis of insulin by pepsin: a process leading to the release of a highly amyloidogenic two chain 'H-fragment'. The H-fragment retains several cleavage sites for pepsin. However, under favorable conditions H-monomers rapidly self-assemble into proteolysis-resistant amyloid fibrils whose composition provides snapshots of early and intermediate stages of the proteolysis. In this work, we report a remarkable agreement of experimentally determined and simulation-predicted cleavage sites on different stages of the proteolysis. Prediction of cleavage sites was based on the comprehensive analysis of the docking interactions from direct simulation of coupled folding and binding of insulin (or its cleaved derivatives) to pepsin. The most frequent interactions were found to be between the pepsin's active site, or its direct vicinity, and the experimentally determined insulin cleavage sites, which suggest that the docking interactions govern the proteolytic process.

Protocols for All-Atom Reconstruction and High-Resolution Refinement of Protein-Peptide Complex Structures

Structural characterizations of protein-peptide complexes may require further improvements. These may include reconstruction of missing atoms and/or structure optimization leading to higher accuracy models. In this work, we describe a workflow that generates accurate structural models of peptide-protein complexes starting from protein-peptide models in C-alpha representation generated using CABS-dock molecular docking. First, protein-peptide models are reconstructed from their C-alpha traces to all-atom representation using MODELLER. Next, they are refined using Rosetta FlexPepDock. The described workflow allows for reliable all-atom reconstruction of CABS-dock models and their further improvement to high-resolution models.