DeepFold: enhancing protein structure prediction through optimized loss functions, improved template features, and re-optimized energy function.

MOTIVATION: Predicting protein structures with high accuracy is a critical challenge for the broad community of life sciences and industry. Despite progress made by deep neural networks like AlphaFold2, there is a need for further improvements in the quality of detailed structures, such as side-chains, along with protein backbone structures. RESULTS: Building upon the successes of AlphaFold2, the modifications we made include changing the losses of side-chain torsion angles and frame aligned point error, adding loss functions for side chain confidence and secondary structure prediction, and replacing template feature generation with a new alignment method based on conditional random fields. We also performed re-optimization by conformational space annealing using a molecular mechanics energy function which integrates the potential energies obtained from distogram and side-chain prediction. In the CASP15 blind test for single protein and domain modeling (109 domains), DeepFold ranked fourth among 132 groups with improvements in the details of the structure in terms of backbone, side-chain, and Molprobity. In terms of protein backbone accuracy, DeepFold achieved a median GDT-TS score of 88.64 compared with 85.88 of AlphaFold2. For TBM-easy/hard targets, DeepFold ranked at the top based on Z-scores for GDT-TS. This shows its practical value to the structural biology community, which demands highly accurate structures. In addition, a thorough analysis of 55 domains from 39 targets with publicly available structures indicates that DeepFold shows superior side-chain accuracy and Molprobity scores among the top-performing groups. AVAILABILITY AND IMPLEMENTATION: DeepFold tools are open-source software available at https://github.com/newtonjoo/deepfold.

Correction: Aggregation or phase separation can be induced in highly charged proteins by small charged biomolecules.

Correction for 'Aggregation or phase separation can be induced in highly charged proteins by small charged biomolecules' by Minchae Kang et al., Soft Matter, 2022, 18, 3313-3317, https://doi.org/10.1039/D2SM00384H.

DNA sequence and methylation prescribe the inside-out conformational dynamics and bending energetics of DNA minicircles.

Eukaryotic genome and methylome encode DNA fragments' propensity to form nucleosome particles. Although the mechanical properties of DNA possibly orchestrate such encoding, the definite link between 'omics' and DNA energetics has remained elusive. Here, we bridge the divide by examining the sequence-dependent energetics of highly bent DNA. Molecular dynamics simulations of 42 intact DNA minicircles reveal that each DNA minicircle undergoes inside-out conformational transitions with the most likely configuration uniquely prescribed by the nucleotide sequence and methylation of DNA. The minicircles' local geometry consists of straight segments connected by sharp bends compressing the DNA's inward-facing major groove. Such an uneven distribution of the bending stress favors minimum free energy configurations that avoid stiff base pair sequences at inward-facing major grooves. Analysis of the minicircles' inside-out free energy landscapes yields a discrete worm-like chain model of bent DNA energetics that accurately account for its nucleotide sequence and methylation. Experimentally measuring the dependence of the DNA looping time on the DNA sequence validates the model. When applied to a nucleosome-like DNA configuration, the model quantitatively reproduces yeast and human genomes' nucleosome occupancy. Further analyses of the genome-wide chromatin structure data suggest that DNA bending energetics is a fundamental determinant of genome architecture.

Aggregation or phase separation can be induced in highly charged proteins by small charged biomolecules.

Protein phase separation in biological systems has captured the attention of scientists in the last decade; however, the main mechanism underlying protein phase separation in cells remains unclear. Biologists, physicists, and chemists have all tried to understand this important biological phenomenon, each using their own unique techniques and language. Each subject has its advantages in explaining protein phase separation; however, in this study, we find that the chemical language of molecular structure is the key to explaining the mechanism underlying protein phase separation. Using fluroescence microscopy and molecular dynamics, this study identifies small multivalently charged biomolecules, such as nucleoside triphosphate (negatively charged) and polyamine (positively charged), as important drivers of phase separation of highly charged proteins in cells.

Sequence-dependent DNA condensation as a driving force of DNA phase separation.

The physical properties of DNA have been suggested to play a central role in spatio-temporal organization of eukaryotic chromosomes. Experimental correlations have been established between the local nucleotide content of DNA and the frequency of inter- and intra-chromosomal contacts but the underlying physical mechanism remains unknown. Here, we combine fluorescence resonance energy transfer (FRET) measurements, precipitation assays, and molecular dynamics simulations to characterize the effect of DNA nucleotide content, sequence, and methylation on inter-DNA association and its correlation with DNA looping. First, we show that the strength of DNA condensation mediated by poly-lysine peptides as a reduced model of histone tails depends on the DNA's global nucleotide content but also on the local nucleotide sequence, which turns out to be qualitatively same as the condensation by spermine. Next, we show that the presence and spatial arrangement of C5 methyl groups determines the strength of inter-DNA attraction, partially explaining why RNA resists condensation. Interestingly, multi-color single molecule FRET measurements reveal strong anti-correlation between DNA looping and DNA-DNA association, suggesting that a common biophysical mechanism underlies them. We propose that the differential affinity between DNA regions of varying sequence pattern may drive the phase separation of chromatin into chromosomal subdomains.

Hierarchical Self-Assembly of Poly-Pseudorotaxanes into Artificial Microtubules.

Hierarchical self-assembly of building blocks over multiple length scales is ubiquitous in living organisms. Microtubules are one of the principal cellular components formed by hierarchical self-assembly of nanometer-sized tubulin heterodimers into protofilaments, which then associate to form micron-length-scale, multi-stranded tubes. This peculiar biological process is now mimicked with a fully synthetic molecule, which forms a 1:1 host-guest complex with cucurbit[7]uril as a globular building block, and then polymerizes into linear poly-pseudorotaxanes that associate laterally with each other in a self-shape-complementary manner to form a tubular structure with a length over tens of micrometers. Molecular dynamic simulations suggest that the tubular assembly consists of eight poly-pseudorotaxanes that wind together to form a 4.5 nm wide multi-stranded tubule.

Rational Design and Construction of Hierarchical Superstructures Using Shape-Persistent Organic Cages: Porphyrin Box-Based Metallosupramolecular Assemblies.

We report a new approach to building hierarchical superstructures using a shape-persistent porous organic cage, which acts as a premade secondary building unit, and coordination chemistry. To illustrate the principle, a zinc-metalated porphyrin box (Zn-PB), a corner-truncated cubic porous cage, was connected by suitable dipyridyl terminated bridging ligands to construct PB-based hierarchical superstructures (PSSs). The PSSs were stabilized not only by the coordination bonds between Zn ions and bipyridyl-terminated ligands but also by π-π interactions between the corners of the Zn-PB units. By varying the length of the linker, we identified an optimum range of the linker length for construction of PSSs. The PSSs have large void volumes and an extrinsic surface area compared to the parent PBs, which can be exploited for the selective encapsulation and interior functionalization of the PSSs for various applications, including catalysis. We observed that singlet oxygen induced synthesis of the natural product, juglone, is more efficiently catalyzed by PSS-1 than its constituent component Zn-PB.

New tricks for old dogs: improving the accuracy of biomolecular force fields by pair-specific corrections to non-bonded interactions.

In contrast to ordinary polymers, the vast majority of biological macromolecules adopt highly ordered three-dimensional structures that define their functions. The key to folding of a biopolymer into a unique 3D structure or to an assembly of several biopolymers into a functional unit is a delicate balance between the attractive and repulsive forces that also makes such self-assembly reversible under physiological conditions. The all-atom molecular dynamics (MD) method has emerged as a powerful tool for studies of individual biomolecules and their functional assemblies, encompassing systems of ever increasing complexity. However, advances in parallel computing technology have outpaced the development of the underlying theoretical models-the molecular force fields, pushing the MD method into an untested territory. Recent tests of the MD method have found the most commonly used molecular force fields to be out of balance, overestimating attractive interactions between charged and hydrophobic groups, which can promote artificial aggregation in MD simulations of multi-component protein, nucleic acid, and lipid systems. One route towards improving the force fields is through the NBFIX corrections method, in which the intermolecular forces are calibrated against experimentally measured quantities such as osmotic pressure by making atom pair-specific adjustments to the non-bonded interactions. In this article, we review development of the NBFIX (Non-Bonded FIX) corrections to the AMBER and CHARMM force fields and discuss their implications for MD simulations of electrolyte solutions, dense DNA systems, Holliday junctions, protein folding, and lipid bilayer membranes.

The structure and intermolecular forces of DNA condensates.

Spontaneous assembly of DNA molecules into compact structures is ubiquitous in biological systems. Experiment has shown that polycations can turn electrostatic self-repulsion of DNA into attraction, yet the physical mechanism of DNA condensation has remained elusive. Here, we report the results of atomistic molecular dynamics simulations that elucidated the microscopic structure of dense DNA assemblies and the physics of interactions that makes such assemblies possible. Reproducing the setup of the DNA condensation experiments, we measured the internal pressure of DNA arrays as a function of the DNA-DNA distance, showing a quantitative agreement between the results of our simulations and the experimental data. Analysis of the MD trajectories determined the DNA-DNA force in a DNA condensate to be pairwise, the DNA condensation to be driven by electrostatics of polycations and not hydration, and the concentration of bridging cations, not adsorbed cations, to determine the magnitude and the sign of the DNA-DNA force. Finally, our simulations quantitatively characterized the orientational correlations of DNA in DNA arrays as well as diffusive motion of DNA and cations.

De novo reconstruction of DNA origami structures through atomistic molecular dynamics simulation.

The DNA origami method has brought nanometer-precision fabrication to molecular biology labs, offering myriads of potential applications in the fields of synthetic biology, medicine, molecular computation, etc. Advancing the method further requires controlling self-assembly down to the atomic scale. Here we demonstrate a computational method that allows the equilibrium structure of a large, complex DNA origami object to be determined to atomic resolution. Through direct comparison with the results of cryo-electron microscopy, we demonstrate de novo reconstruction of a 4.7 megadalton pointer structure by means of fully atomistic molecular dynamics simulations. Furthermore, we show that elastic network-guided simulations performed without solvent can yield similar accuracy at a fraction of the computational cost, making this method an attractive approach for prototyping and validation of self-assembled DNA nanostructures.

Mono-allyloxylated Cucurbit[7]uril Acts as an Unconventional Amphiphile To Form Light-Responsive Vesicles.

Serendipitously, mono-allyloxylated cucurbit[7]uril (AO1 CB[7]) was discovered to act as an unconventional amphiphile which self-assembles into light-responsive vesicles (AO1 CB[7]VC) in water. Although the mono-allyloxy group, directly tethered on the periphery of CB[7], is much shorter (C4) than the hydrophobic tails of conventional amphiphiles, it played an important role in vesicle formation. Light-activated transformation of the allyloxy group by conjugation with glutathione was exploited as a remote tool to disrupt the vesicle. The vesicle showed on-demand release of cargo upon irradiation by a laser, after they were internalized into cancer cells. This result demonstrated the potential of AO1 CB[7]VC as a new type of light-responsive intracellular delivery vehicle for the release of therapeutic cargo, within cells, on demand.

Ion Channels Made from a Single Membrane-Spanning DNA Duplex.

Because of their hollow interior, transmembrane channels are capable of opening up pathways for ions across lipid membranes of living cells. Here, we demonstrate ion conduction induced by a single DNA duplex that lacks a hollow central channel. Decorated with six porpyrin-tags, our duplex is designed to span lipid membranes. Combining electrophysiology measurements with all-atom molecular dynamics simulations, we elucidate the microscopic conductance pathway. Ions flow at the DNA-lipid interface as the lipid head groups tilt toward the amphiphilic duplex forming a toroidal pore filled with water and ions. Ionic current traces produced by the DNA-lipid channel show well-defined insertion steps, closures, and gating similar to those observed for traditional protein channels or synthetic pores. Ionic conductances obtained through simulations and experiments are in excellent quantitative agreement. The conductance mechanism realized here with the smallest possible DNA-based ion channel offers a route to design a new class of synthetic ion channels with maximum simplicity.

Improved model of hydrated calcium ion for molecular dynamics simulations using classical biomolecular force fields.

Calcium ions (Ca(2+) ) play key roles in various fundamental biological processes such as cell signaling and brain function. Molecular dynamics (MD) simulations have been used to study such interactions, however, the accuracy of the Ca(2+) models provided by the standard MD force fields has not been rigorously tested. Here, we assess the performance of the Ca(2+) models from the most popular classical force fields AMBER and CHARMM by computing the osmotic pressure of model compounds and the free energy of DNA-DNA interactions. In the simulations performed using the two standard models, Ca(2+) ions are seen to form artificial clusters with chloride, acetate, and phosphate species; the osmotic pressure of CaAc2 and CaCl2 solutions is a small fraction of the experimental values for both force fields. Using the standard parameterization of Ca(2+) ions in the simulations of Ca(2+) -mediated DNA-DNA interactions leads to qualitatively wrong outcomes: both AMBER and CHARMM simulations suggest strong inter-DNA attraction whereas, in experiment, DNA molecules repel one another. The artificial attraction of Ca(2+) to DNA phosphate is strong enough to affect the direction of the electric field-driven translocation of DNA through a solid-state nanopore. To address these shortcomings of the standard Ca(2+) model, we introduce a custom model of a hydrated Ca(2+) ion and show that using our model brings the results of the above MD simulations in quantitative agreement with experiment. Our improved model of Ca(2+) can be readily applied to MD simulations of various biomolecular systems, including nucleic acids, proteins and lipid bilayer membranes. © 2016 Wiley Periodicals, Inc. Biopolymers 105: 752-763, 2016.

In situ structure and dynamics of DNA origami determined through molecular dynamics simulations.

The DNA origami method permits folding of long single-stranded DNA into complex 3D structures with subnanometer precision. Transmission electron microscopy, atomic force microscopy, and recently cryo-EM tomography have been used to characterize the properties of such DNA origami objects, however their microscopic structures and dynamics have remained unknown. Here, we report the results of all-atom molecular dynamics simulations that characterized the structural and mechanical properties of DNA origami objects in unprecedented microscopic detail. When simulated in an aqueous environment, the structures of DNA origami objects depart from their idealized targets as a result of steric, electrostatic, and solvent-mediated forces. Whereas the global structural features of such relaxed conformations conform to the target designs, local deformations are abundant and vary in magnitude along the structures. In contrast to their free-solution conformation, the Holliday junctions in the DNA origami structures adopt a left-handed antiparallel conformation. We find the DNA origami structures undergo considerable temporal fluctuations on both local and global scales. Analysis of such structural fluctuations reveals the local mechanical properties of the DNA origami objects. The lattice type of the structures considerably affects global mechanical properties such as bending rigidity. Our study demonstrates the potential of all-atom molecular dynamics simulations to play a considerable role in future development of the DNA origami field by providing accurate, quantitative assessment of local and global structural and mechanical properties of DNA origami objects.

Three-dimensional stress field around a membrane protein: atomistic and coarse-grained simulation analysis of gramicidin A.

Using both atomistic and coarse-grained (CG) models, we compute the three-dimensional stress field around a gramicidin A (gA) dimer in lipid bilayers that feature different degrees of negative hydrophobic mismatch. The general trends in the computed stress field are similar at the atomistic and CG levels, supporting the use of the CG model for analyzing the mechanical features of protein/lipid/water interfaces. The calculations reveal that the stress field near the protein-lipid interface exhibits a layered structure with both significant repulsive and attractive regions, with the magnitude of the stress reaching 1000 bar in certain regions. Analysis of density profiles and stress field distributions helps highlight the Trp residues at the protein/membrane/water interface as mechanical anchors, suggesting that similar analysis is useful for identifying tension sensors in other membrane proteins, especially membrane proteins involved in mechanosensation. This work fosters a connection between microscopic and continuum mechanics models for proteins in complex environments and makes it possible to test the validity of assumptions commonly made in continuum mechanics models for membrane mediated processes. For example, using the calculated stress field, we estimate the free energy of membrane deformation induced by the hydrophobic mismatch, and the results for regions beyond the annular lipids are in general consistent with relevant experimental data and previous theoretical estimates using elasticity theory. On the other hand, the assumptions of homogeneous material properties for the membrane and a bilayer thickness at the protein/lipid interface being independent of lipid type (e.g., tail length) appear to be oversimplified, highlighting the importance of annular lipids of membrane proteins. Finally, the stress field analysis makes it clear that the effect of even rather severe hydrophobic mismatch propagates to only about two to three lipid layers, thus putting a limit on the range of cooperativity between membrane proteins in crowded cellular membranes.

Membrane-mediated protein-protein interactions and connection to elastic models: a coarse-grained simulation analysis of gramicidin A association.

To further foster the connection between particle based and continuum mechanics models for membrane mediated biological processes, we carried out coarse-grained (CG) simulations of gramicidin A (gA) dimer association and analyzed the results based on the combination of potential of mean force (PMF) and stress field calculations. Similar to previous studies, we observe that the association of gA dimers depends critically on the degree of hydrophobic mismatch, with the estimated binding free energy of >10 kcal/mol in a distearoylphosphatidylcholine bilayer. Qualitative trends in the computed PMF can be understood based on the stress field distributions near a single gA dimer and between a pair of gA dimers. For example, the small PMF barrier, which is ∼1 kcal/mol independent of lipid type, can be captured nearly quantitatively by considering membrane deformation energy associated with the region confined by two gA dimers. However, the PMF well depth is reproduced poorly by a simple continuum model that only considers membrane deformation energy beyond the annular lipids. Analysis of lipid orientation, configuration entropy, and stress distribution suggests that the annular lipids make a significant contribution to the association of two gA dimers. These results highlight the importance of explicitly considering contributions from annular lipids when constructing approximate models to study processes that involve a significant reorganization of lipids near proteins, such as protein-protein association and protein insertion into biomembranes. Finally, large-scale CG simulations indicate that multiple gA dimers also form clusters, although the preferred topology depends on the protein concentration. Even at high protein concentrations, every gA dimer requires contact to lipid hydrocarbons to some degree, and at most three to four proteins are in contact with each gA dimer; this observation highlights another aspect of the importance of interactions between proteins and annular lipids.

A comparison of coarse-grained and continuum models for membrane bending in lipid bilayer fusion pores.

To establish the validity of continuum mechanics models quantitatively for the analysis of membrane remodeling processes, we compare the shape and energies of the membrane fusion pore predicted by coarse-grained (MARTINI) and continuum mechanics models. The results at these distinct levels of resolution give surprisingly consistent descriptions for the shape of the fusion pore, and the deviation between the continuum and coarse-grained models becomes notable only when the radius of curvature approaches the thickness of a monolayer. Although slow relaxation beyond microseconds is observed in different perturbative simulations, the key structural features (e.g., dimension and shape of the fusion pore near the pore center) are consistent among independent simulations. These observations provide solid support for the use of coarse-grained and continuum models in the analysis of membrane remodeling. The combined coarse-grained and continuum analysis confirms the recent prediction of continuum models that the fusion pore is a metastable structure and that its optimal shape is neither toroidal nor catenoidal. Moreover, our results help reveal a new, to our knowledge, bowing feature in which the bilayers close to the pore axis separate more from one another than those at greater distances from the pore axis; bowing helps reduce the curvature and therefore stabilizes the fusion pore structure. The spread of the bilayer deformations over distances of hundreds of nanometers and the substantial reduction in energy of fusion pore formation provided by this spread indicate that membrane fusion can be enhanced by allowing a larger area of membrane to participate and be deformed.

Conformational changes in the human Cx43/GJA1 gap junction channel visualized using cryo-EM.

Connexin family proteins assemble into hexameric hemichannels in the cell membrane. The hemichannels dock together between two adjacent membranes to form gap junction intercellular channels (GJIChs). We report the cryo-electron microscopy structures of Cx43 GJICh, revealing the dynamic equilibrium state of various channel conformations in detergents and lipid nanodiscs. We identify three different N-terminal helix conformations of Cx43-gate-covering (GCN), pore-lining (PLN), and flexible intermediate (FIN)-that are randomly distributed in purified GJICh particles. The conformational equilibrium shifts to GCN by cholesteryl hemisuccinates and to PLN by C-terminal truncations and at varying pH. While GJIChs that mainly comprise GCN protomers are occluded by lipids, those containing conformationally heterogeneous protomers show markedly different pore sizes. We observe an α-to-π-helix transition in the first transmembrane helix, which creates a side opening to the membrane in the FIN and PLN conformations. This study provides basic structural information to understand the mechanisms of action and regulation of Cx43 GJICh.

Cryo-EM structures of human Cx36/GJD2 neuronal gap junction channel.

Connexin 36 (Cx36) is responsible for signal transmission in electrical synapses by forming interneuronal gap junctions. Despite the critical role of Cx36 in normal brain function, the molecular architecture of the Cx36 gap junction channel (GJC) is unknown. Here, we determine cryo-electron microscopy structures of Cx36 GJC at 2.2-3.6 Å resolutions, revealing a dynamic equilibrium between its closed and open states. In the closed state, channel pores are obstructed by lipids, while N-terminal helices (NTHs) are excluded from the pore. In the open state with pore-lining NTHs, the pore is more acidic than those in Cx26 and Cx46/50 GJCs, explaining its strong cation selectivity. The conformational change during channel opening also includes the α-to-π-helix transition of the first transmembrane helix, which weakens the protomer-protomer interaction. Our structural analyses provide high resolution information on the conformational flexibility of Cx36 GJC and suggest a potential role of lipids in the channel gating.