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.

Structural basis of transcription: RNA polymerase II substrate binding and metal coordination using a free-electron laser.

Catalysis and translocation of multisubunit DNA-directed RNA polymerases underlie all cellular mRNA synthesis. RNA polymerase II (Pol II) synthesizes eukaryotic pre-mRNAs from a DNA template strand buried in its active site. Structural details of catalysis at near-atomic resolution and precise arrangement of key active site components have been elusive. Here, we present the free-electron laser (FEL) structures of a matched ATP-bound Pol II and the hyperactive Rpb1 T834P bridge helix (BH) mutant at the highest resolution to date. The radiation-damage-free FEL structures reveal the full active site interaction network, including the trigger loop (TL) in the closed conformation, bonafide occupancy of both site A and B Mg2+, and, more importantly, a putative third (site C) Mg2+ analogous to that described for some DNA polymerases but not observed previously for cellular RNA polymerases. Molecular dynamics (MD) simulations of the structures indicate that the third Mg2+ is coordinated and stabilized at its observed position. TL residues provide half of the substrate binding pocket while multiple TL/BH interactions induce conformational changes that could allow translocation upon substrate hydrolysis. Consistent with TL/BH communication, a FEL structure and MD simulations of the T834P mutant reveal rearrangement of some active site interactions supporting potential plasticity in active site function and long-distance effects on both the width of the central channel and TL conformation, likely underlying its increased elongation rate at the expense of fidelity.

In Vitro Encapsulation of Functionally Active Abiotic Photosensitizers Inside a Bacterial Microcompartment Shell.

Bacterial microcompartments (BMCs) are self-assembling, selectively permeable protein shells that encapsulate enzymes to enhance catalytic efficiency of segments of metabolic pathways through means of confinement. The modular nature of BMC shells' structure and assembly enables programming of shell permeability and underscores their promise in biotechnology engineering efforts for applications in industry, medicine, and clean energy. Realizing this potential requires methods for encapsulation of abiotic molecules, which have been developed here for the first time. We report in vitro cargo loading of BMC shells with ruthenium photosensitizers (RuPS) by two approaches─one involving site-specific covalent labeling and the other driven by diffusion, requiring no specific interactions between cargo molecules and shell proteins. The highly stable shells retain encapsulated cargo over 1 week without egress and preserve RuPS photophysical activity. This study is an important foundation for further work that will converge biological BMC architecture with synthetic chemistry to facilitate biohybrid photocatalysis.

De Novo Synthesis and Structural Elucidation of CDR-H3 Loop Mimics.

The binding affinity of antibodies to specific antigens stems from a remarkably broad repertoire of hypervariable loops known as complementarity-determining regions (CDRs). While recognizing the pivotal role of the heavy-chain 3 CDRs (CDR-H3s) in maximizing antibody-antigen affinity and specificity, the key structural determinants responsible for their adaptability to diverse loop sequences, lengths, and noncanonical structures are hitherto unknown. To address this question, we achieved a de novo synthesis of bulged CDR-H3 mimics excised from their full antibody context. CD and NMR data revealed that these stable standalone ß-hairpin scaffolds are well-folded and retain many of the native bulge CDR-H3 features in water. In particular, the tryptophan residue, highly conserved across CDR-H3 sequences, was found to extend the kinked base of these ß-bulges through a combination of stabilizing intramolecular hydrogen bond and CH/π interaction. The structural ensemble consistent with our NMR observations exposed the dynamic nature of residues at the base of the loop, suggesting that ß-bulges act as molecular hinges connecting the rigid stem to the more flexible loops of CDR-H3s. We anticipate that this deeper structural understanding of CDR-H3s will lay the foundation to inform the design of antibody drugs broadly and engineer novel CDR-H3 peptide scaffolds as therapeutics.

Stability and deformation of biomolecular condensates under the action of shear flow.

Biomolecular condensates play a key role in cytoplasmic compartmentalization and cell functioning. Despite extensive research on the physico-chemical, thermodynamic, or crowding aspects of the formation and stabilization of the condensates, one less studied feature is the role of external perturbative fluid flow. In fact, in living cells, shear stress may arise from streaming or active transport processes. Here, we investigate how biomolecular condensates are deformed under different types of shear flows. We first model Couette flow perturbations via two-way coupling between the condensate dynamics and fluid flow by deploying Lattice Boltzmann Molecular Dynamics. We then show that a simplified approach where the shear flow acts as a static perturbation (one-way coupling) reproduces the main features of the condensate deformation and dynamics as a function of the shear rate. With this approach, which can be easily implemented in molecular dynamics simulations, we analyze the behavior of biomolecular condensates described through residue-based coarse-grained models, including intrinsically disordered proteins and protein/RNA mixtures. At lower shear rates, the fluid triggers the deformation of the condensate (spherical to oblated object), while at higher shear rates, it becomes extremely deformed (oblated or elongated object). At very high shear rates, the condensates are fragmented. We also compare how condensates of different sizes and composition respond to shear perturbation, and how their internal structure is altered by external flow. Finally, we consider the Poiseuille flow that realistically models the behavior in microfluidic devices in order to suggest potential experimental designs for investigating fluid perturbations in vitro.

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.

Robust Measurements of n-Point Correlation Functions of Driven-Dissipative Quantum Systems on a Digital Quantum Computer.

We propose and demonstrate a unified hierarchical method to measure n-point correlation functions that can be applied to driven, dissipative, or otherwise open or nonequilibrium quantum systems. In this method, the time evolution of the system is repeatedly interrupted by interacting an ancilla qubit with the system through a controlled operation, and measuring the ancilla immediately afterward. We discuss the robustness of this method as compared to other ancilla-based interferometric techniques (such as the Hadamard test), and highlight its advantages for near-term quantum simulations of open quantum systems. We implement the method on a quantum computer in order to measure single-particle Green's functions of a driven-dissipative fermionic system. This Letter shows that dynamical correlation functions for driven-dissipative systems can be robustly measured with near-term quantum computers.

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.

Non-Abelian topological order and anyons on a trapped-ion processor.

Non-Abelian topological order is a coveted state of matter with remarkable properties, including quasiparticles that can remember the sequence in which they are exchanged1-4. These anyonic excitations are promising building blocks of fault-tolerant quantum computers5,6. However, despite extensive efforts, non-Abelian topological order and its excitations have remained elusive, unlike the simpler quasiparticles or defects in Abelian topological order. Here we present the realization of non-Abelian topological order in the wavefunction prepared in a quantum processor and demonstrate control of its anyons. Using an adaptive circuit on Quantinuum's H2 trapped-ion quantum processor, we create the ground-state wavefunction of D4 topological order on a kagome lattice of 27 qubits, with fidelity per site exceeding 98.4 per cent. By creating and moving anyons along Borromean rings in spacetime, anyon interferometry detects an intrinsically non-Abelian braiding process. Furthermore, tunnelling non-Abelions around a torus creates all 22 ground states, as well as an excited state with a single anyon-a peculiar feature of non-Abelian topological order. This work illustrates the counterintuitive nature of non-Abelions and enables their study in quantum devices.

One bead per residue can describe all-atom protein structures.

Atomistic resolution is the standard for high-resolution biomolecular structures, but experimental structural data are often at lower resolution. Coarse-grained models are also used extensively in computational studies to reach biologically relevant spatial and temporal scales. This study explores the use of advanced machine learning networks for reconstructing atomistic models from reduced representations. The main finding is that a single bead per amino acid residue allows construction of accurate and stereochemically realistic all-atom structures with minimal loss of information. This suggests that lower resolution representations of proteins may be sufficient for many applications when combined with a machine learning framework that encodes knowledge from known structures. Practical applications include the rapid addition of atomistic detail to low-resolution structures from experiment or computational coarse-grained models. The application of rapid, deterministic all-atom reconstruction within multi-scale frameworks is further demonstrated with a rapid protocol for the generation of accurate models from cryo-EM densities close to experimental structures.

Structural basis of transcription: RNA Polymerase II substrate binding and metal coordination at 3.0 Å using a free-electron laser.

Catalysis and translocation of multi-subunit DNA-directed RNA polymerases underlie all cellular mRNA synthesis. RNA polymerase II (Pol II) synthesizes eukaryotic pre-mRNAs from a DNA template strand buried in its active site. Structural details of catalysis at near atomic resolution and precise arrangement of key active site components have been elusive. Here we present the free electron laser (FEL) structure of a matched ATP-bound Pol II, revealing the full active site interaction network at the highest resolution to date, including the trigger loop (TL) in the closed conformation, bonafide occupancy of both site A and B Mg2+, and a putative third (site C) Mg2+ analogous to that described for some DNA polymerases but not observed previously for cellular RNA polymerases. Molecular dynamics (MD) simulations of the structure indicate that the third Mg2+ is coordinated and stabilized at its observed position. TL residues provide half of the substrate binding pocket while multiple TL/bridge helix (BH) interactions induce conformational changes that could propel translocation upon substrate hydrolysis. Consistent with TL/BH communication, a FEL structure and MD simulations of the hyperactive Rpb1 T834P bridge helix mutant reveals rearrangement of some active site interactions supporting potential plasticity in active site function and long-distance effects on both the width of the central channel and TL conformation, likely underlying its increased elongation rate at the expense of fidelity.

The effect of polymer length in liquid-liquid phase separation.

Understanding the thermodynamics that drive liquid-liquid phase separation (LLPS) is quite important given the number of diverse biomolecular systems undergoing this phenomenon. Many studies have focused on condensates of long polymers, but very few systems of short-polymer condensates have been observed and studied. Here, we study a short-polymer system of various lengths of poly-adenine RNA and peptides formed by the RGRGG sequence repeats to understand the underlying thermodynamics of LLPS. Using the recently developed COCOMO coarse-grained (CG) model, we predicted condensates for lengths as short as 5-10 residues, which was then confirmed by experiment, making this one of the smallest LLPS systems yet observed. A free-energy model reveals that the length dependence of condensation is driven primarily by entropy of confinement. The simplicity of this system will provide the basis for understanding more biologically realistic systems.

Varying molecular interactions explain aspects of crowder-dependent enzyme function of a viral protease.

Biochemical processes in cells, including enzyme-catalyzed reactions, occur in crowded conditions with various background macromolecules occupying up to 40% of cytoplasm's volume. Viral enzymes in the host cell also encounter such crowded conditions as they often function at the endoplasmic reticulum membranes. We focus on an enzyme encoded by the hepatitis C virus, the NS3/4A protease, which is crucial for viral replication. We have previously found experimentally that synthetic crowders, polyethylene glycol (PEG) and branched polysucrose (Ficoll), differently affect the kinetic parameters of peptide hydrolysis catalyzed by NS3/4A. To gain understanding of the reasons for such behavior, we perform atomistic molecular dynamics simulations of NS3/4A in the presence of either PEG or Ficoll crowders and with and without the peptide substrates. We find that both crowder types make nanosecond long contacts with the protease and slow down its diffusion. However, they also affect the enzyme structural dynamics; crowders induce functionally relevant helical structures in the disordered parts of the protease cofactor, NS4A, with the PEG effect being more pronounced. Overall, PEG interactions with NS3/4A are slightly stronger but Ficoll forms more hydrogen bonds with NS3. The crowders also interact with substrates; we find that the substrate diffusion is reduced much more in the presence of PEG than Ficoll. However, contrary to NS3, the substrate interacts more strongly with Ficoll than with PEG crowders, with the substrate diffusion being similar to crowder diffusion. Importantly, crowders also affect the substrate-enzyme interactions. We observe that both PEG and Ficoll enhance the presence of substrates near the active site, especially near catalytic H57 but Ficoll crowders increase substrate binding more than PEG molecules.

Characterization of RNA polymerase II trigger loop mutations using molecular dynamics simulations and machine learning.

Catalysis and fidelity of multisubunit RNA polymerases rely on a highly conserved active site domain called the trigger loop (TL), which achieves roles in transcription through conformational changes and interaction with NTP substrates. The mutations of TL residues cause distinct effects on catalysis including hypo- and hyperactivity and altered fidelity. We applied molecular dynamics simulation (MD) and machine learning (ML) techniques to characterize TL mutations in the Saccharomyces cerevisiae RNA Polymerase II (Pol II) system. We did so to determine relationships between individual mutations and phenotypes and to associate phenotypes with MD simulated structural alterations. Using fitness values of mutants under various stress conditions, we modeled phenotypes along a spectrum of continual values. We found that ML could predict the phenotypes with 0.68 R2 correlation from amino acid sequences alone. It was more difficult to incorporate MD data to improve predictions from machine learning, presumably because MD data is too noisy and possibly incomplete to directly infer functional phenotypes. However, a variational auto-encoder model based on the MD data allowed the clustering of mutants with different phenotypes based on structural details. Overall, we found that a subset of loss-of-function (LOF) and lethal mutations tended to increase distances of TL residues to the NTP substrate, while another subset of LOF and lethal substitutions tended to confer an increase in distances between TL and bridge helix (BH). In contrast, some of the gain-of-function (GOF) mutants appear to cause disruption of hydrophobic contacts among TL and nearby helices.

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.

Modeling Concentration-dependent Phase Separation Processes Involving Peptides and RNA via Residue-Based Coarse-Graining.

Biomolecular condensation, especially liquid-liquid phase separation, is an important physical process with relevance for a number of different aspects of biological functions. Key questions of what drives such condensation, especially in terms of molecular composition, can be addressed via computer simulations, but the development of computationally efficient yet physically realistic models has been challenging. Here, the coarse-grained model COCOMO is introduced that balances the polymer behavior of peptides and RNA chains with their propensity to phase separate as a function of composition and concentration. COCOMO is a residue-based model that combines bonded terms with short- and long-range terms, including a Debye-Hückel solvation term. The model is highly predictive of experimental data on phase-separating model systems. It is also computationally efficient and can reach the spatial and temporal scales on which biomolecular condensation is observed with moderate computational resources.

Molecular Dynamics Simulations of Rhodamine B Zwitterion Diffusion in Polyelectrolyte Solutions.

Polyelectrolytes continue to find wide interest and application in science and engineering, including areas such as water purification, drug delivery, and multilayer thin films. We have been interested in the dynamics of small molecules in a variety of polyelectrolyte (PE) environments; in this paper, we report simulations and analysis of the small dye molecule rhodamine B (RB) in several very simple polyelectrolyte solutions. Translational diffusion of the RB zwitterion has been measured in fully atomistic, 2 µs long molecular dynamics simulations in four different polyelectrolyte solutions. Two solutions contain the common polyanion sodium poly(styrene sulfonate) (PSS), one with a 30-mer chain and the other with 10 trimers. The other two solutions contain the common polycation poly(allyldimethylammonium) chloride (PDDA), one with two 15-mers and the other with 10 trimers. RB diffusion was also simulated in several polymer-free solutions to verify its known experimental value for the translational diffusion coefficient, DRB, of 4.7 × 10-6 cm2/s at 300 K. RB diffusion was slowed in all four simulated PE solutions, but to varying degrees. DRB values of 3.07 × 10-6 and 3.22 × 10-6 cm2/s were found in PSS 30-mer and PSS trimer solutions, respectively, whereas PDDA 15-mer and trimer solutions yielded values of 2.19 × 10-6 and 3.34 × 10-6 cm2/s. Significant associations between RB and the PEs were analyzed and interpreted via a two-state diffusion model (bound and free diffusion) that describes the data well. Crowder size effects and anomalous diffusion were also analyzed. Finally, RB translation along the polyelectrolytes during association was characterized.

Characterizing Transient Protein-Protein Interactions by Trp-Cys Quenching and Computer Simulations.

Transient protein-protein interactions occur frequently under the crowded conditions encountered in biological environments. Tryptophan-cysteine quenching is introduced as an experimental approach with minimal labeling for characterizing such interactions between proteins due to its sensitivity to nano- to microsecond dynamics on subnanometer length scales. The experiments are paired with computational modeling at different resolutions including fully atomistic molecular dynamics simulations for interpretation of the experimental observables and to gain molecular-level insights. This approach is applied to model systems, villin variants and the drkN SH3 domain, in the presence of protein G crowders. It is demonstrated that Trp-Cys quenching experiments can differentiate between overall attractive and repulsive interactions between different proteins, and they can discern variations in interaction preferences at different protein surface locations. The close integration between experiment and simulations also provides an opportunity to evaluate different molecular force fields for the simulation of concentrated protein solutions.

Entanglement from Tensor Networks on a Trapped-Ion Quantum Computer.

The ability to selectively measure, initialize, and reuse qubits during a quantum circuit enables a mapping of the spatial structure of certain tensor-network states onto the dynamics of quantum circuits, thereby achieving dramatic resource savings when simulating quantum systems with limited entanglement. We experimentally demonstrate a significant benefit of this approach to quantum simulation: the entanglement structure of an infinite system-specifically the half-chain entanglement spectrum-is conveniently encoded within a small register of "bond qubits" and can be extracted with relative ease. Using Honeywell's model H0 quantum computer equipped with selective midcircuit measurement and reset, we quantitatively determine the near-critical entanglement entropy of a correlated spin chain directly in the thermodynamic limit and show that its phase transition becomes quickly resolved upon expanding the bond-qubit register.

Multi-state modeling of G-protein coupled receptors at experimental accuracy.

The family of G-protein coupled receptors (GPCRs) is one of the largest protein families in the human genome. GPCRs transduct chemical signals from extracellular to intracellular regions via a conformational switch between active and inactive states upon ligand binding. While experimental structures of GPCRs remain limited, high-accuracy computational predictions are now possible with AlphaFold2. However, AlphaFold2 only predicts one state and is biased toward either the active or inactive conformation depending on the GPCR class. Here, a multi-state prediction protocol is introduced that extends AlphaFold2 to predict either active or inactive states at very high accuracy using state-annotated templated GPCR databases. The predicted models accurately capture the main structural changes upon activation of the GPCR at the atomic level. For most of the benchmarked GPCRs (10 out of 15), models in the active and inactive states were closer to their corresponding activation state structures. Median RMSDs of the transmembrane regions were 1.12 Å and 1.41 Å for the active and inactive state models, respectively. The models were more suitable for protein-ligand docking than the original AlphaFold2 models and template-based models. Finally, our prediction protocol predicted accurate GPCR structures and GPCR-peptide complex structures in GPCR Dock 2021, a blind GPCR-ligand complex modeling competition. We expect that high accuracy GPCR models in both activation states will promote understanding in GPCR activation mechanisms and drug discovery for GPCRs. At the time, the new protocol paves the way towards capturing the dynamics of proteins at high-accuracy via machine-learning methods.