Browse Articles

How good is generative diffusion model for enhanced sampling of protein conformations across scales and in all-atom resolution?

Molecular dynamics (MD) simulations are fundamental for probing the structural dynamics of biomolecules, yet their efficiency is limited by the high computational cost of exploring long-timescale events. Generative machine learning (ML) models, particularly the Denoising Diffusion Probabilistic Model (DDPM), offer an emerging strategy to enhance conformational sampling. In this study, we evaluate the capabilities and limitations of DDPM in generating atomistically accurate conformational ensembles across proteins of varying size and structural order, ranging from the 20-residue folded Trp-cage and 58-residue BPTI to the 83-residue intrinsically disordered region Ash1 and the 140-residue intrinsically disordered protein α-Synuclein. Training DDPM on relatively short MD trajectories using both torsion angle and all-atom coordinate data, we demonstrate that it can reproduce key structural features such as secondary structure, radius of gyration, and contact maps, while effectively sampling sparsely populated regions of the conformational landscape. Notably, DDPM can also generate novel conformations, including transitions not explicitly observed in the training data. However, the model occasionally overlooks low-probability regions and may produce conformers with unclear physical relevance, warranting independent validation. These limitations are particularly evident in flexible systems such as IDPs. Overall, this work benchmarks DDPM as a viable tool for augmenting MD simulations, offering enhanced sampling with significant computational savings, while noting its limitations in capturing low-populated conformers. At the same time, it highlights the importance of rigorous validation and thoughtful interpretation when deploying generative models in computational biophysics.

Gate-controlled two-stage Kondo effect in a metal-free polycyclic aromatic hydrocarbon diradical molecular switch: A DFT+NRG insight

Strong correlation behaviors play a pivotal role in realizing novel, high-performance, low-power nanoscale electronic devices. In this study, we theoretically design a molecular switch based on a metal-free polycyclic aromatic hydrocarbon diradical molecule dominated by strong electron-electron repulsions. By establishing a customized simulation framework that combines density functional theory with the numerical renormalization group method, and with the aid of a side-coupled two-orbital Anderson model, we systematically simulate the quantum transport governed by various kinds of Kondo effects. When the central energy level sweeps upward, the linear conductance tends to reach its unitary limit in the low-temperature regime, driven by the broadening two-stage Kondo effect window. As the side energy increases, a quantum phase transition from the two-stage Kondo effect to the spin-1/2 Kondo effect emerges, resulting in a stable, full conductance plateau. Furthermore, by tuning the exchange coupling between the central and the side orbitals, we simulate the impact of precisely adjusting the intramolecular dihedral angle of the molecule on the transport properties. The linear conductance exhibits a transition from zero to the unitary limit as the molecule smoothly evolves from the antiferromagnetic to the ferromagnetic regime. These findings provide valuable insights into the dynamical and thermodynamical properties of complex polycyclic aromatic hydrocarbon systems. Our suggested “ab initio + model calculation” theoretical framework may offer a promising methodology for exploring the complex Kondo physics in real magnetic nanosystems.

douka: A universal platform of data assimilation for materials modeling

A large-scale, general-purpose data assimilation (DA) platform for materials modeling, douka, was developed and applied to nonlinear materials models. The platform demonstrated its effectiveness in estimating physical properties that cannot be directly obtained from observed data. Furthermore, it enables state estimation with quantified uncertainty, thereby providing researchers with a new aspect for analyzing the underlying physical process and guiding future model refinements. DA was successfully performed using experimental images of oxygen evolution reaction at a water electrolysis electrode, enabling the estimation of oxygen gas injection velocity and bubble contact angle. Furthermore, large-scale ensemble DA was conducted on the supercomputer Fugaku, achieving state estimation with up to 8192 ensemble members. The results confirmed that runtime scaling for the prediction step follows the weak scaling law, ensuring computational efficiency even with increased ensemble sizes. These findings highlight the potential of douka as a new approach for data-driven materials science, integrating experimental data with numerical simulation.

dMSGB-IE: Computational mutational scanning for (de)methylation thermodynamics

The (de)methylation regulates the functional interactions between the unstructured N-terminal of histones and other globular proteins. The multistate behavior of methyl-substitution makes the situation complex; for example, being mono-methylated, di-methylated, or tri-methylated. As a pivotal epigenetic marker, understanding its thermodynamic impact on protein-protein binding is crucial for the elucidation of the regulation mechanism of epigenetic modifications on target genes. To this aim, in this work, we present a cost-effective free energy technique named computational (de)methylation scanning with generalized Born and interaction entropy (dMSGB-IE). Our regime is built on implicit-solvent-based end-point free energy techniques and provides an efficient route to access the (de)methylation-induced affinity change with a screening power comparable to costlier alchemical free energy calculations. We first use a batch of histone-reader recognition protein-protein complexes as illustrative cases, showing the capabilities and reliabilities of dMSGB-IE. Then, we augment the method with the integrative structure prediction tool AlphaFold 3, providing a fully computational workflow for fast estimation of (de)methylation free energies. Based on a batch of testing systems, we validate the practical applicability and highlight the predictive power of the promising integrative modeling workflow.

Unveiling the interplay of electronic and phononic excitations in laser-induced oxygen activation on Ru(0001)

Understanding laser-induced dynamics on metal surfaces poses significant challenges due to the intricate interplay between electronic and phononic degrees of freedom, which evolve on distinct timescales. In this study, we introduce a machine learning-accelerated approach to molecular dynamics simulations that incorporates anisotropic electronic friction, providing deeper insights into these complex processes. Our framework extends the accessible time and length scales for nonadiabatic dynamics simulations, enabling a detailed investigation of the laser-induced activation of oxygen on the Ru(0001) surface. Statistical analysis reveals that strong electronic excitation dominates the first 800 fs after laser exposure. Beyond this timescale, energy deposited by electronic excitation continues to drive oxygen activation, while phonons, although always present as a dissipation channel, play a weaker role by buffering energy loss and redistributing kinetic energy among vibrational modes. The observed non-linear yield–fluence relationship, described by Y ∼ Fn, underscores the pivotal role of electronic excitation. In addition, we identify the z-direction as the key activation mode for oxygen diffusion, with the exponent of the power law representing the quantized energy required for this process. This approach significantly accelerates dynamic simulations while offering valuable insights into the interplay between electronic and phononic excitations during laser-induced oxygen activation on Ru(0001).

Dynamics of stiff filaments in size-polydisperse hard sphere fluids

The dynamics of a rod-shaped stiff filament (formed by connecting beads) embedded in a size-polydisperse fluid of soft repulsive spheres were investigated using molecular dynamics simulations with a focus on how the degree of size-polydispersity, characterized by the polydispersity index (δ), affects the dynamics in this model heterogeneous system. Polydispersity of the fluid and strong coupling of rotational and translational motions of the rods are two of the various hurdles in interpreting experimental results in complex fluid environments. Furthermore, the influence of volume fraction, ϕ, and absolute free volume, Vfree, which changes inherently with δ, on the dynamics is not adequately discussed in the literature. Thus, we investigate the dynamical behavior of the rods under two conditions: (i) constant pressure (in which ϕ changes with δ) and (ii) constant ϕ. Under constant pressure, it is observed that the rotational relaxation time and, hence, the rotational diffusion constant, DR, vary with rod length, l, as DR ∼ l−α, where the value of exponent α increases from ∼3.0 to 3.2 while varying δ from 0% to 40%. It is observed that the effect of increasing ϕ dominates over the effect of increasing Vfree. Moreover, hydrodynamic interactions among beads within a rod contribute minimally to rotational dynamics, although partial hydrodynamic screening is observed for center-of-mass motion. Meanwhile, for fixed ϕ systems, increasing δ results in increasing Vfree and thus enhances tracer diffusion, a trend opposite to that observed under constant pressure.

Interaction of gas molecules with cyclo[N]carbon: Size dependence and atomic doping modulation

The unique electronic properties of cyclo[N]carbon have attracted considerable attention due to their potential applications in gas storage and sensing technologies. This work employed density functional theory (DFT) and DLPNO-CCSD(T) to investigate the molecular adsorption characteristics of cyclo[N]carbon (N = 12, 14, and 16) with various gas molecules. It is interesting that the adsorption strength of cyclo[N]carbon for gases increases as the size of cyclo[N]carbon increases, with polar molecules demonstrating stronger interactions than nonpolar ones. Local energy decomposition analysis at the high-end DLPNO-CCSD(T) level of theory reveals that London dispersion forces significantly contribute to adsorption stability. By incorporating the Mg2 dimer in two-layer C16, a stable Mg22+@(C16)22− complex is formed, and the encapsulation of divalent cations considerably enhances the gas molecule adsorption performance. This study provides valuable insights into the adsorption properties of cyclo[N]carbon, which could be crucial for advancements in next-generation molecular devices.

Independent switchable atomic silver quantum transistor via potential-driven surface reconstruction

Atomic-scale quantum conductance switches based on metallic quantum point contacts allow controlled binary switching of the electrical current between a conducting “on state” and a non-conducting “off state” via an independent gate electrode. Although silver-based quantum switches operate via electrochemical conductance modulation, the underlying atomic-scale mechanisms remain unclear. In this study, we employed density functional theory combined with the computational hydrogen electrode (CHE) framework to investigate nitrate anion adsorption on Ag(100), Ag(111), and Ag(511) surfaces under varying electrochemical potentials. The stable configurations of nitrate adsorption on the electrode surfaces as a function of electrode potential are determined using the grand canonical approach within the CHE framework, revealing the surface reconstruction of Ag(511) analogous to the formation of nascent Ag–NO3 complexes. The thermodynamic analysis indicates that the critical phase transition occurs in agreement with the experimental switching threshold. Electronic structure calculations reveal metallic to semiconducting transitions in the reconstructed Ag–NO3 complex layers, elucidating the mechanism of conductance modulation. The “off state” corresponds to insulating Ag–NO3 surface complexes, while the “on state” arises from metallic Ag–Ag bridging at lower nitrate coverage under lower control voltage. This study establishes a general framework for anion-mediated quantum switching in transition metals and provides a design for nanoscale electrochemical devices.

A new generation of effective core potentials: Selected lanthanides and heavy elements II

We present a new set of correlation-consistent effective core potentials (ccECPs) for selected heavy s, p, d, and f-block elements significant in materials science and chemistry (Rb, Sr, Cs, Ba, In, Sb, Pb, Ru, Cd, La, Ce, and Eu). The ccECPs are designed using minimal Gaussian parameterization to achieve smooth and bounded potentials. They are expressed as a combination of averaged relativistic effective potentials (AREPs) and effective spin–orbit terms, developed within a relativistic coupled-cluster framework. The optimization is driven by correlated all-electron (AE) atomic spectra, norm-conservation, and spin–orbit splittings, with considerations for plane wave cutoffs to ensure accuracy and viability across various electronic configurations. The transferability of these ccECPs is validated through testing on molecular oxides and hydrides, emphasizing discrepancies in molecular binding energies across a spectrum of bond lengths and electronic environments. The ccECPs demonstrate excellent agreement with AE reference calculations, attaining chemical accuracy in bond dissociation energies and equilibrium bond lengths, even in systems characterized by substantial relativistic and correlation effects. These ccECPs provide an accurate and transferable framework for valence-only calculations.

Pressure-induced structural transitions of diamond (100) surfaces

Despite extensive research conducted on the structural transitions of crystalline solids under pressure, the transitions occurring on the solid surfaces that transmit the pressure have been relatively neglected. Here, we investigate the pressure-induced structural transitions of the diamond (100) surface using molecular dynamics simulations combined with the volume-Constant Pressure Molecular Dynamics method for finite systems. Eight possible dimerized configurations were identified through an exhaustive method, considering both translational and rotational symmetries, which in turn define eight diamond (100) surfaces. These surfaces are nearly degenerate in energy at zero pressure, but their energy differences become larger under high external pressure. At finite temperatures, the increasing pressure induces graphitization of the surfaces. The transition pressures differ among the various surfaces. By calculating the free energies of the surfaces, we determined the most stable surfaces at various pressures and temperatures and constructed a schematic P–T “phase diagram” to illustrate the stability competition and structural transitions of the surfaces. This study provides a theoretical basis for the efficient utilization of diamond under high pressure and offers insights into the surface properties of materials under extreme conditions.

Benchmarking distinguishable cluster methods to platinum standard CCSDT(Q) non-covalent interaction energies in the A24 dataset

Recent disagreement between state-of-the-art quantum chemical methods, coupled cluster with single, double, and perturbative triples excitations and fixed-node diffusion Monte Carlo, calls for a systematic examination of possible sources of error within both methodological approaches. Coupled cluster (CC) theory is systematically improvable toward the exact solution of the Schrödinger equation; however, it is very quickly limited by the computational cost of the calculation. Therefore, it has become imperative to develop low-cost methods that are able to reproduce CC results beyond the CC theory with single, double, and perturbative triples [CCSD(T)] level of theory. Here, the distinguishable cluster (DC)-CCSDT and singular value decomposed (SVD)-DC-CCSDT methods are examined for their fidelity to the CCSDT(Q) correlation interaction energies for the A24 dataset and are shown to outperform CCSDT and CCSD(T). Furthermore, with (T)-based corrections of the SVD approximation, the SVD-DC-CCSDT method becomes an accurate and relatively low-cost tool for the calculation of previously intractable post-CCSD(T) energies in atomic orbital basis sets of unprecedented size.

Equations of state and excess entropy of repulsive inverse power particle potential fluids with variable stiffness

Analytic expressions for the equation of state in terms of the compressibility factor, Z, and excess entropy, sex, of inverse power (IP) potential fluids are derived and parameterized using molecular dynamics simulation data. The IP pair potential is ϕ(r) ∼ r−n, where r is the pair separation and n is an exponent that governs the steepness of the potential. A number of parameterizations of the dependence of Z on number density and n are proposed and compared. These include multiparameter global series expansion fits and more simple formulas based on the very soft (small n) and hard sphere particle representation limits. The excess entropy can be represented over the whole fluid and n > 3+ range well by a single analytic expression with n-dependent parameters. It is shown that sex is a concave function of density for n greater than about 6 and convex for smaller values of the exponent. The excess entropy varies with n for a constant value of the density normalized by its freezing point value.

Erratum: “Machine learning of kinetic energy densities with target and feature smoothing: Better results with fewer training data” [J. Chem. Phys. 159, 234115 (2023)]

On the <i>in</i>existence of a “splay(-bend)” nematic phase

On the basis of the application of the (Onsager) second-virial density functional theory to an artificial system that is so designed as to be the best promoter of a “splay(-bend)” nematic phase, it is argued that this “modulated” nematic phase cannot exist.

Advances in unveiling water’s molecular mysteries

The strain-stiffening critical exponents in polymer networks and their universality

Disordered athermal biopolymer materials, such as collagen networks that constitute a major component in extracellular matrices and various connective tissues, are initially soft and compliant but stiffen dramatically under strain. Such network materials are topologically sub-isostatic and feature strong rigidity scale separation between the bending and stretching response of the constituent polymer fibers. Recently, a comprehensive scaling theory of the athermal strain-stiffening phase transition has been developed, providing predictions for all mean-field critical exponents characterizing the transition in terms of the distance to the critical strain and of the small rigidity scales ratio. Here, we employ large-scale computer simulations, at and away from criticality, to test the analytic predictions. We find that all numerical critical exponents are in quantitative agreement with the analytically predicted ones. Moreover, we find that all predicted mean-field exponents remain valid whether the driving strain is shear, i.e., volume-preserving, or dilation, and independent of the degree of the network’s sub-isostaticity, thus establishing the universality of the strain-stiffening phase transition with respect to the symmetry of the driving strain and the network’s topology.

An automated QM/MM average protein electrostatic configuration approach for flavoproteins: APEC-F 2.0

Flavoproteins are a ubiquitous class of redox proteins, enzymes, and photoreceptors that derive their versatility from the flavin cofactor—a prosthetic group that serves as the main locus of their spectral, photophysical, and (photo)chemical properties. It is thus common for computational modeling of flavoproteins to employ a hybrid approach that treats the flavin quantum mechanically and the remaining atoms classically. Such quantum mechanical/molecular mechanical (QM/MM) methods have proven powerful for studying flavoproteins so far, but users are often faced with a choice between treating the flavin electronic structure with ab initio wave function methods or using more approximate methods that allow for more extensive sampling of the protein dynamics. Herein, we present APEC-F 2.0, an automated QM/MM workflow that uses several open-source software packages to construct QM/MM models of flavoproteins. Exploiting the rigidity of flavin’s tricyclic isoalloxazine ring, the APEC approach iteratively optimizes flavin’s geometry in a static MM environment that represents a dynamic protein using a superposition of configurations generated from molecular dynamics. The automation of the code enables the systematic construction of QM/MM models using a common protocol and is suitable for comparing flavin’s spectral, electronic, and chemical properties in different redox, protonation, or excited states in a wide range of flavoproteins.

A double-U approach to more accurate metal adsorption energies on ceria

An empirical approach based on density functional theory (DFT)+U is proposed to provide adsorption energies for atoms of transition metals adsorbed on ceria that are more accurate than those found in the existing DFT literature. It involves applying an intermediate Hubbard U value to the Ce 4f states and another U value to the d states of the metal adatom that is optimized for each metal to match its calculated bulk cohesive energy with the experimental value. The results compare favorably with Campbell’s calorimetric measurements of metal adsorption heats on CeO2(111). Thus, this approach produces metal adsorption energies on ceria that are more accurate in both the single-atom and the bulk limits. Furthermore, it preserves a common reference for Ce so that calculations involving different metals adsorbed on ceria, e.g., bimetallic and multimetallic clusters, can be analyzed on the same footing.

Modulation of spin states and electronic excitation via molecular doping with Fe(II)-porphyrin in 2D gallium nitride

2D gallium nitride possesses distinctive electronic states, making it ideal for future optoelectronic devices because of the quantum confinement and enhanced many-body interactions inherent in its atomically thin form. This study explores the impact of molecular doping with Fe(II)-porphyrin (FeP) on these characteristics using first-principle calculations. Contact of the magnetic center Fe with the nitrogen site causes a 17% decrease in the energy barrier for the transition from intermediate spin (S = 1) to high spin (S = 2) state, highlighting the sensitivity of spin dynamics to doping sites. Molecular diffusion barriers increase by 9.6 kJ/mol upon spin transition, suggesting that the spin state influences molecular mobility within the material. Exploring spectral functions reveals that FeP doping introduces spin-dependent molecule levels within the bandgap, which may play a role in electron–hole separation and spin injection. In addition, we show that the molecule–substrate coupling lowers the exciton binding energy by 0.1 eV, with further reduction during spin transitions. This weakening is attributed to increased electron mobility, quantified by static polarizability. These results indicate that the molecular spin state can control electronic excitations within substrate materials, presenting a promising strategy for designing spintronic devices.

Exciton diffusion in MoS2 monolayer from first-principles molecular dynamics

First-principles modeling of exciton dynamics coupled with lattice vibrations is important for understanding exciton mobility, which is crucial for various applications. In order to shed light on such coupled exciton–lattice dynamics, in this paper, we use a restricted open-shell Kohn–Sham approach, which is a computationally efficient method for electronic structure calculations in the lowest excited states. Within this framework, we analyze the correlated electron–hole dynamics of a bright exciton in the 1H–MoS2 monolayer in real space at different temperatures and obtain the exciton diffusion rate that is in reasonable agreement with the experimental findings.